1、Research on Key Technologies of Bistatic Sensing and PrototypeVerification1 Overview.11.1 Background on Bistatic Sensing.11.2 Comparative Analysis of Bistatic Sensing and Monostatic Sensing Performance.52 Bistatic Sensing System Architecture.102.1 Introduction to ISAC Architecture.102.2 Sensing Serv

2、ice Procedures.122.3 Measurement Quantities Definition and Sensing Information Transmission.182.3.1 Definition of Sensing Measurement Quantities.182.3.2 Protocol Stack of Sensing Information Transmission.212.3.3 Reporting of Measurement Results.223 Bistatic Sensing Key Technologies.253.1 Signal Desi

3、gn.253.1.1 Sequence Design.253.1.2 Resource Mapping Schemes.283.1.3 Sensing and Communication Signal Resource Multiplexing.343.2 Estimation of Bistatic Sensing Parameters.393.2.1 Basic Methods of Spectrum Estimation.393.2.2 Position Information Calculation for Bistatic Sensing.423.2.3 NLOS-Based Sen

4、sing Method.443.3 Non-Ideal Factors Cancellation.473.3.1 Hardware Non-Ideal Factors and Impacts.473.3.2 Hardware Non-Ideal Factor Cancellation Schemes.483.3.3 Bistatic Clutter Cancellation Schemes.523.4 Beam Management and Precoding.543.4.1 Beam Management for Bistatic Sensing.543.4.2 Digital Precod

5、ing Schemes of Bistatic Sensing.604 Simulation Evaluation and Prototype Verification.634.1 Simulation Evaluation Methods.634.1.1 Bistatic Sensing Channel Modeling.634.1.2 Bistatic Sensing Simulation Evaluation Methods.654.2 Prototype Verification.694.2.1 Collaborative Positioning Among Base Stations

6、.694.2.2 Respiratory Monitoring Through Bistatic Sensing.704.2.3 Coordinated Multi-Point Trajectory Tracking Through Bistatic Sensing.744.2.4 Passive Target Positioning Through Bistatic Sensing.764.2.5 Bistatic Test Based on LOS and NLOS Synchronization Calibration.784.2.6 Action Detection Through B

7、istatic Sensing.815 Summary and Outlook.83References.84Abbreviations.87Contributors to the White Paper.901/901Overview1.1 Background on Bistatic SensingIn the field of radar technology,systems are categorized into monostatic radars or bistatic radarsbased on whether the transmitting and receiving an

8、tennas are co-located or separate.For bistaticradar,the baseline distance between the transmitter and receiver is typically comparable to theradars operational range1.Bistatic radar can operate by receiving signals from either dedicatedcooperative transmitters or non-cooperative transmitters designe

9、d for other purposes.The latteroperational paradigms is refered to as passive radar,passive coherent location(PCL)radar,passive surveillance radar,or non-cooperative illuminator radar2,which is a specific case ofbistatic radar that employs electromagnetic wave detection theories and signal processin

10、gtechniques to exploit third-party emissions,enabling target detection,localization,tracking,andidentification.Bistatic radar finds extensive applications spanning military,civilian,and scientificdomains,serving critical functions in diverse application scenarios such as aerial/maritime targetdetect

11、ion and tracking,meteorological observation and prediction,and traffic surveillance.The characteristics of bistatic radar are closely tied to its geometry,particularly the bistatictriangle formed by the transmitter,target,and receiver.The plane containing this bistatic triangleis referred to as the

12、bistatic plane,as illustrated in Figure 1-1.The distance between the transmitterand receiver is referred to as the baseline distance or simply the baseline.The angle formedbetween the line from the transmitter to the target and the line from the receiver to the target iscalled the bistatic angle,whi

13、ch also serves to highlight the performance differences betweenmonostatic and bistatic radars.The range information directly measured by the bistatic radarreceiver typically represents the sum of the distances from the target to the transmitter and fromthe target to the receiver,minus the baseline d

14、istance,known as the bistatic range difference.When the baseline distance is known,the sum of the distances from the target to the transmitterand from the target to the receiver can be calculated.In bistatic radar,calculating the distancefrom the target to the receiver or transmitter or determining

15、the targets position involves solvingthe bistatic triangle,which typically requires obtaining the relative position information betweenthe transmitter and the receiver.2/90RxTargetTxRTBistatic TriangleBaselineRRFigure 1-1 Geometric Relationships in Bistatic RadarAccording to the definition of bistat

16、ic radar,the sensing mode in which the transmitting andreceiving antennas are separated is collectively referred to as the bistatic sensing mode.Based onthe current mobile communication network architecture,six sensing modes can be supported:(1)base stations self-transmit and self-receive,(2)base st

17、ation A transmits while base station Breceives;(3)base station transmits and terminal receives;(4)terminal transmits and base stationreceives;(5)terminals self-transmit and self-receive,and(6)terminal A transmits while terminalB receives.Modes(1)and(5)are classified as monostatic sensing modes,while

18、 the remainingfour modes are categorized as bistatic sensing modes.The advantages and disadvantages of thesesensing modes are summarized in Table 1-1.Table 1-1 Comparison of Advantages and Disadvantages of Different Sensing ModesSensing modesAdvantagesDisadvantages(1)Basestationsself-transmitandself

19、-receiveNo transmitter-receiversynchronization issuesBase station requires full duplexcapability(2)Basestation Atransmitswhile basestation BreceivesDoes not require full-duplexcapability,relatively lowrequirements for CP length andunambiguous rangingSynchronization errors can impactsensing performan

20、ce,lowerprobability of LOS path acquisitioncompared to the first sensing mode,complex pairing3/90(3)Basestationtransmits andterminalreceivesCan reuse or enhance existingsignals,does not require full-duplexcapability,large number ofterminals,and the sensing coverageperformance can be improved byselec

21、ting a suitable terminal device,relatively low requirements for CPlength and unambiguous rangingIssues such as terminal position,orientation,velocity,body blockageand synchronization errors can affectsensing performance,requires certainterminal antenna specifications forangular measurements(4)Termin

22、altransmits andbase stationreceivesCan reuse or enhance existingsignals,does not require full-duplexcapability,large number ofterminals,and the sensing coverageperformance can be improved byselecting a suitable terminal device,relatively low requirements for CPlength and unambiguous rangingIssues su

23、ch as terminal position,orientation,velocity,body blockageand synchronization errors can affectsensing performance,lowtransmission power,limited number ofsupported transmit antennas(5)Terminalsself-transmitandself-receiveNo transmitter-receiversynchronization issues,supportssensing without network c

24、overageTerminal requiresfull-duplexcapability,low transmission power,limited number of supported transmitantennas(6)Terminal Atransmitswhile terminalB receivesCan reuse or enhance existingsignals,does not require full-duplexcapability,supports sensing withoutnetwork coverageIssues such as terminal p

25、osition,orientation,velocity,body blockageand synchronization errors can affectsensing performance,requires certainterminal antenna specifications forangular measurements,lowtransmission power,limited number ofsupported transmit antennasBistatic sensing demonstrates numerous advantages over traditio

26、nal monostatic sensing,serving asa crucial foundation for achieving high accuracy and wide-coverage sensing.Firstly,bistaticsensing offers more available sensing paths.By selecting the optimal node with a line-of-sight(LOS)path from multiple candidate receiving nodes,it effectively reduces sensing b

27、lind spots.Compared to the LOS probability in monostatic sensing,bistatic sensing significantly enhancesthe coverage probability of LOS paths by increasing the number of nodes,thereby improving4/90target detection capabilities in complex environments.Secondly,bistatic sensing utilizes higherscatteri

28、ng intensity to receive target echoes from different directions,significantly enhancing theaverage scattering intensity of signals.In addition,through the coordination of multiple bases,thespatial sensing coverage has been expanded.For a moving target,bistatic sensing enablescontinuous sensing throu

29、gh dynamic node selection and efficient handover,ensuring accuratetracking of the target throughout its entire activity range.In monostatic sensing,the receiver is affected by strong self-interference from transmissionleakage,which can lead to receiver saturation or reduce the dynamic range of the s

30、ensing signal3.Methods to reduce self-interference include two aspects:interference isolation and interferencecancellation.Interference isolation aims to minimize the transmission leakage received by thereceiver,achieved by spatial separation to reduce received leakage through increased distance,orb

31、y using metal or absorbing materials to provide RF shielding between the transmitter andreceiver.Interference cancellation is used to mitigate the impact of the received transmissionleakage,and is mainly implemented by weighted cancellation of the known transmitted signal inthe analog or digital dom

32、ain45.Bistatic sensing does not have the issue of self-interference,eliminating the need for corresponding hardware modifications.Moreover,bistatic sensingdemonstrates greater resistance to single-point interference due to its diverse signal sources,whereas monostatic sensing is more vulnerable to d

33、irectional interference.Target recognitioncapability is another advantage of bistatic sensing.By utilizing multi-angle observation,itcaptures more comprehensive scattering characteristics of targets,thereby improving recognitionaccuracy.In contrast,monostatic sensing may result in a higher false rec

34、ognition rate due to itslimited feature information of targets.Coordination gain is another advantage of bistatic sensing.Through the cooperation of bistatic andmulti-static modes,not only enhances the signal-to-noise ratio(SNR)and target detectionprobability,but also effectively reduces random erro

35、rs through geometric gain.At the same time,the redundant information provided by multiple nodes can be used to eliminate faulty nodes,enhancing the reliability,stability,and overall restoration capability of the sensing system in athree-dimensional environment.Bistatic sensing also offers flexible i

36、nterference coordinationcapabilities.By dynamically designating certain nodes as collaborative receivers,it effectivelymitigates various interference issues,including direct signal interference and mutual interference,while optimizing overall communication network performance and enhancing sensing a

37、ccuracy.However,bistatic sensing also faces several challenges.First and foremost,clutter suppressionposes a significant challenge,particularly when the echo signal is weak,necessitating effectivealgorithms to mitigate clutter interference and enhance sensing accuracy.Secondly,precisesynchronization

38、 is essential,as any timing synchronization error between the transmitting andreceiving nodes directly impacts sensing positioning accuracy.For instance,a synchronizationerror of 10 nanoseconds can result in a positioning error of up to 3 meters.Non-line-of-sight(NLOS)identification and utilization

39、is another core challenge of bistatic sensing,particularly in5/90determining how to recognize and use NLOS sensing paths when blocking occurs.Furthermore,when sensing targets are in motion,efficient mechanisms are needed for node selection andswitching to ensure the continuity and accuracy of sensin

40、g tasks.Increasing system complexity isalso a significant challenge,as it requires effective synchronization and data fusion between basestations,which adds to the overall system complexity.The deployment and maintenance costs ofmultiple base stations are also higher than those of a monostatic syste

41、m.Additionally,the bistaticsystem demands a communication network that offers high bandwidth and low latency betweenbase stations.The complexity of signal processing has also increased,requiring moresophisticated algorithms to handle multi-source data,which in turn raises the computationalburden.1.2

42、 Comparative Analysis of Bistatic Sensing and Monostatic SensingPerformanceSensing Coverage:There are differences in coverage performance between bistatic and monostatic sensing.Considering the impact of noise,the coverage of monostatic sensing can be defined as a circulararea,whereas the coverage r

43、ange of bistatic sensing is defined as a Cassini oval.The bistaticsensing range equation can be expressed as6 max=22243 min(1)where,and represent the distance from the target to the transmitter and receiverrespectively;represents the transmitting power,and represent the gain of thetransmitting and r

44、eceiving antennas respectively,is the wavelength,is the radar crosssection(RCS)of the target,and and represent the path propagation factor from thetransmitter or receiver to the target respectively.is the Boltzmann constant,is the receivernoise temperature,is the receiver noise bandwidth,minis the s

45、ignal-to-noise ratio(SNR)required for target detection,and and are the transmitter and receiver systemlosses respectively.Let =22243,the bistatic sensing SNR can be expressed as =22,and then thebistatic constant SNR contours are obtained as shown in Figure 1-2,where =304,and isthe baseline distance.

46、6/90Figure 1-2 Bistatic Sensing Constant SNR ContoursIt can be observed that,depending on the relationship between the baseline distance and thebistatic distance,the maximum coverage area of bistatic sensing can be approximately an ellipse,a Cassini oval,or split into two elliptical parts.For bistat

47、ic sensing,a better SNR can be achievedin areas closer to the transmitter or receiver.Therefore,bistatic sensing can enhance coverageperformance compared to monostatic sensing by optimally selecting the transmitter or receiver.This is particularly beneficial for sensing tasks involving terminals,as

48、the greater number andflexibility of terminal locations can be leveraged to enhance sensing performance.Figure 1-3displays the simulation results of the signal-to-interference-plus-noise ratio(SINR)for monostaticsensing with base stations self-transmission and reception,and for bistatic sensing,wher

49、e basestations transmit while user equipments receive the sensing signal in an Urban Micro(Umi)scenario.The sensing SINR represents the ratio of the signal power from the path or clusterreflected by the sensing target to the interference and noise.For bistatic sensing,it is assumed thatthe distance

50、between the sensing target and the UE does not exceed 10 meters.Simulation resultsindicate that bistatic sensing achieves better sensing coverage performance than monostaticsensing in a mobile cellular network architecture,as its receiving node is closer to the sensingtarget.Additionally,bistatic se

51、nsing offers more observation angles than monostatic sensing,providing significant benefits for applications such as environmental reconstruction and targetrecognition.7/90Figure 1-3 Sensing SINR Simulation Results in an Umi ScenarioSensing Measurement Range and Resolution:Coverage performance is di

52、rectly related to the sensing signal-to-noise ratio(SNR)or sensingSINR,whereas the sensing unambiguous range and resolution performance depend on signalresource configuration.Compared to monostatic sensing,the unambiguous range/velocity andrange/velocity resolution for bistatic sensing are influence

53、d not only by the signal configurationbut also by the bistatic geometry.Table 1-2 illustrates the relationship between the resourceconfiguration of monostatic and bistatic sensing signals and their unambiguous measurementrange and resolution performance7.Table 1-2 Sensing Performance and Calculation

54、 FormulasSensing PerformanceCalculation FormulasInfluencing FactorsUnambiguousrangeMonostaticsensing=2Sampling interval infrequency domainBistaticsensing=22Sampling interval infrequency domain,bistaticangleUnambiguousvelocityMonostaticsensing=4Wavelength,sensing signalperiodBistaticsensing=42Wavelen

55、gth,sensing signalperiod and bistatic angleUnambiguous angle=sin12Wavelength,antenna spacingRangeresolutionMonostaticsensing=2Bandwidth8/90Bistaticsensing=2 2Bandwidth,bistatic angleVelocityresolutionMonostaticsensing=2Wavelength,sensing symbolperiod and number ofsensing symbolsBistaticsensing=22Wav

56、elength,sensing symbolperiod,number of sensingsymbols and bistatic angleAngle resolution 0.886Wavelength,array apertureNote:B-signal bandwidth,M-number of pulses,Tr-pulse repetition period,f-sampling intervalin frequency domain,3dB-3db beam width,c-the speed of light,-signal wavelength,D-arrayapertu

57、re,d-antenna spacing,-bistatic angle.In ranging applications,it is crucial to consider both the constraints of the frequency domainresource interval and the limitations imposed by the cyclic prefix(CP)on the measurement range.When the sensing target is distant from the transceiver,the delay of the r

58、eceived sensing echosignal may exceed the CP range,leading to inter-symbol interference and adversely affectingranging performance.For bistatic sensing,range information is typically derived from the relativedelay between the echo signal of the sensing target and the LOS signal.Consequently,compared

59、with monostatic sensing,bistatic sensing has lower requirements for the unambiguous range,including the frequency domain resource spacing and CP length.Blind Spot of Sensing Detection:The sensing system primarily utilizes Doppler detection for moving target identification,wherethe echo signal reflec

60、ted from the target is typically superimposed with a Doppler frequency.However,certain scenarios may arise where either there is no detectable Doppler frequency or thefrequency is minimal,commonly referred to as blind speeds8.In monostatic sensing mode,when the target moves perpendicular to the dire

61、ction of the sensing signal,the delay betweenadjacent sampling points in the time domain remains constant,resulting in no detectable Dopplerfrequency.Similarly,in bistatic sensing mode,if the target follows an oval path with thetransmitting and receiving equipment as foci,the signal time delay stays

62、 unchanged,again leadingto no observable Doppler frequency.Moreover,if the targets movement direction is closelyaligned with the previously mentioned scenarios,the resultant Doppler frequency shiftsuperimposed on the sensing signal due to the targets movement may still be small andchallenging to det

63、ect.9/90In addition to the blind speeds,monostatic and bistatic sensing also exhibit differences in theirrange measurement blind spots.In monostatic sensing mode,when using pulsed signals withtime-division transmission and reception,any echo signal arriving at the receiving antenna beforethe start o

64、f signal reception will not be received,resulting in a close-range blind region8.Inbistatic sensing mode,range information is derived by measuring the relative delay between theecho signal from the sensing target and the LOS signal.Therefore,when the sensing target islocated along the line connectin

65、g the transmitter and receiver,its echo signal and the LOS signalarrive nearly simultaneously.Consequently,the time delay difference is either negligible ornonexistent,making it difficult to accurately measure the distance to the sensing target.Thisregion is known as the forward scattering region of

66、 the bistatic sensing system.In this area,thesystem lacks the ability to directly determine the targets position due to insufficient rangeresolution,and parameters related to the targets motion could be estimated using information suchas echo Doppler9.Sensing Resource Overhead:Introducing sensing ca

67、pabilities into communication systems typically leads to resourcecompetition with communication services,potentially degrading data rates.For monostaticsensing,additional hardware modifications are required to facilitate the reception of echo signals,and when employing continuous wave for sensing,fu

68、ll-duplex capability is required for effectiveself-interference suppression.For bistatic sensing,the communication transceiver link can bemultiplexed,allowing the communication signal to be utilized for sensing or enhanced for sensingpurposes while meeting the requirements of sensing coverage.There

69、is no need to allocatededicated sensing signal resources,and the communication service will not be affected.10/902Bistatic Sensing System Architecture2.1 Introduction to ISAC ArchitectureThe overall architecture of Integrated Sensing and Communication(ISAC)is illustrated in Figure2-1.Key functions r

70、elated to sensing include sensing measurement,sensing data processing,sensing data storage,sensing data application,and sensing control and management.-Sensing measurement refers to the process in which the receiving node acquiressensingmeasurements from the received sensing signal,such as time dela

71、y,Doppler,and angle,etc.via signal processing.In ISAC systems,both UE and radio access network(RAN)possesssensing measurement capabilities.For bistatic sensing,the transmitting and receiving nodesare distinct nodes.For example,a base station may transmit sensing signals while UEreceives them,or vice

72、 versa.Additionally,bistatic sensing scenarios also involve one basestation transmitting sensing signals while another base station receives them,or UE Atransmitting sensing signals while UE B receives them.-Sensing data processing refers to extract target information by further analyzing the aboves

73、ensing measurement data(e.g.,channel response,time delay,Doppler,and angle).In ISACsystems,the sensing data processing is an optional function for UE or RAN.The availabilityof sensing data processing on the UE depends on the UEs willingness or computingcapability for the processing of the sensing me

74、asurement data.For RAN,it depends onwhether the RAN node is allowed to further process the sensing measurement data.Forexample,if the sensing target information obtained from further processing is sensitive,theRAN node is not allowed to perform this kind of processing of sensing data.Correspondingly

75、,the sensing data processing function should be handled in the core network(CN).-Sensing data storage refers to the storage of sensing data in UE,RAN,CN or third-partyservers,based on the sensing requirements and policies of privacy and security.-Sensing data application involves utilizing sensed ta

76、rget information in specific applicationscenarios.Application scenarios of ISAC can be categorized into two types:networkproviding sensing service to application function(also known as NET for Sensing)andsensing-assisted network(also known as Sensing for NET).In the condition of NET forSensing,the s

77、ensing data application is located in CN or UE.In the condition of Sensing forNET,the sensing data application is located in the RAN.-Sensing control and management involves the control and management during the sensingprocess,which includes sensing capability reporting,sensing node selection,config

78、uration ofsensing measurements,and management of sensing data privacy and security.The sensing information includes both sensing signaling and sensing data.Sensing informationtransmission includes air interface transmission and wired interface transmission.The air interfacetransmission of sensing in

79、formation includes the following three possible methods and theircombinations:11/90-Method 1:Transmit sensing information based on the control plane.Sensing signaling anddata are transmitted via the enhancement of existing signaling bearers(such as SRB,MACCE,and UCI formats).-Method 2:Transmit sensi

80、ng information based on the user plane.Sensing signaling and dataare transmitted via the enhancement of existing data bearer(such as DRB).-Method 3:Transmit sensing information based on a newly defined data plane bearer.Thenewly data plane bearer supports flexible prioritization,termination points a

81、nd optimizedconfiguration of protocol stack functions(e.g.,PDCP configuration)based on thetransmission requirements of sensing signaling and data.The wired interface transmission of sensing information includes the following three methods andtheir combinations:-Method 1:Transmit sensing information

82、based on the control plane,and enhance signaling ordata transmission that supports sensing on the existing signaling interface(such as N2 andXn based on SCTP/IP);-Method 2:Transmit sensing information based on the user plane,and enhance signaling ordata transmission that supports sensing on the exis

83、ting user plane interface(such as N3 basedon GTP-U/UDP/IP).-Method 3:Transmit sensing information based on the data plane,and define a new data planeinterface to support sensing signaling or data transmission.The newly defined data planeinterface supports flexible topology based on sensing signaling

84、 or data requirements.12/90Figure 2-1 Architecture of ISAC System2.2 Sensing Service ProceduresThe sensing service procedures encompass sensing node selection,sensing capability reporting,and configuration of sensing measurements.Compared with monostatic sensing,messageinteraction between the transm

85、itter and receiver must also be taken into account for bistaticsensing.a)Sensing node selectionFirst,select the sensing area and mode according to the sensing service requirements.Then,selectbase stations and/or terminals to participate in sensing according to the chosen sensing mode andarea.Only no

86、des authorized to select a sensing node may perform the sensing node selectionprocess.The right to choose a sensing node is closely related to the sensing mode.In certainsensing modes,it is more appropriate for the sensing function(SF)to handle the selection ofsensing nodes,while in others,the base

87、station is better suited for this task.Additionally,a hybridapproach may be employed in some cases.For instance,in the mode of base station transmitting-base station receiving,the SF may select the transmitting base station,which in turn determinesthe receiving base station for the sensing response.

88、13/90The selection of a sensing node requires consideration of several factors,including the mobility ofcandidate nodes,their capabilities to meet sensing service requirements,the presence of a LOSpath,and the state of the candidate nodes,such as whether it is currently engaged in sensing orcommunic

89、ation services,the availability of frequency bands and computing resources.Thesensing node selection procedure should minimize interaction and consist primarily of a selectioninstruction and confirmation.Specifically,the selecting device must confirm with the chosen nodewhether it is willing to assu

90、me the role of sensing node,as illustrated in Figure 2-2.Figure 2-2 Selection Procedure of Sensing NodeCompared to monostatic sensing mode,selecting sensing nodes in bistatic sensing mode is morecomplex.For example,it requires separate consideration of the mobility,capabilities,and states ofboth the

91、 sensing signal-transmitting equipment and the sensing response-receiving equipment.Additionally,scenarios such as re-selection due to changes in the transmitting or receivingequipment must be taken into account.One approach of sensing node selection is to categorize sensing nodes into subsets based

92、 ondifferent sensing modes,such as those where base stations transmit and receive on their own,orthose where base stations transmit while terminals receive.During the node establishment andupdate phases,all subsets must be evaluated to determine suitable sensing modes andcorresponding nodes,resultin

93、g in the final selected subset.Selection criteria for sensing modes and nodes can be tailored for specific scenarios and use cases.For instance,a base station self-transmit and self-receive mode might be primarily utilized,whilea terminal-assisted bistatic sensing mode may be employed at the cell ed

94、ge or in base stationcoverage blind spots,as illustrated in Figure 2-3.For complex scenarios,AI can also be used toassist in the selection of sensing modes and nodes.Each selected subset conducts specific sensingoperations,including configuration,transmission and reception of sensing signals,signalp

95、rocessing,and calculation of sensing results.14/90Figure 2-3 Example of Node Selection for Collaborative Bistatic Sensingb)Sensing capability reportingFirst,sensing capability reporting should be on demand rather than mandatory.It should not bebundled with the UEs communication capability reports.Re

96、porting sensing capabilities withoutan active sensing service may lead to a waste of communication resources and energy.Whenreporting is on demand,sensing capabilities can be transmitted as data without control signaling.Secondly,only equipment authorized to select a sensing node should report sensi

97、ng capabilities.For example,the selection of a sensing node is performed by the SF,base station,or UE,and thenthe sensing nodes capability is correspondingly reported to the SF,base station,or UE.Forbistatic sensing,the equipment responsible for selecting sensing nodes must obtain capabilitiesfrom b

98、oth the transmitting and receiving equipment to make an informed decision.Additionally,candidate nodes can actively report their sensing capabilities based on their role as transmitting orreceiving nodes.For instance,if a node is chosen as a sensing signal transmitter,it should onlyreport its transm

99、ission-related capabilities or equipment status.Similarly,if it acts as a receivingnode,it should report capabilities related to reception.The procedure for requesting and reportingsensing capabilities is illustrated in Figure 2-4,where the reported sensing capabilities includesupported sensing mode

100、s,transmitting and receiving capabilities,sensing measurement methodsfor those modes,and the accuracy of the corresponding measurement methods.Figure 2-4 Sensing Capability Reporting Procedurec)Configuration of sensing measurementThe configuration of sensing measurement focuses on how to execute sen

101、sing operations,including setups for transmitting measurement values and for providing feedback on sensingresults.For bistatic sensing,configurations related to sensing signal transmission are sentexclusively to the transmitting equipment,while configurations for signal reception,measurement,15/90or

102、 feedback on sensing results are directed to the receiving equipment,particularly if the resultsneed to be transmitted externally.For the configuration of sensing measurement,the control equipment of the sensing measurementresources should configure the transmitting resources.For instance,in the sen

103、sing modes of basestation transmitting-UE receiving,UE transmitting-base station receiving,and UE transmitting-UE receiving,the base station has full control over resource allocation,making it responsible fortransmitting the sensing measurement configuration.Therefore,the process of measurement conf

104、iguration based on different modes may include thefollowing cases:For the sensing modes of base station transmitting-UE receiving and UE transmitting-basestation receiving,the base station sends the sensing measurement configuration,as shown inFigure 2-5.Figure 2-5 Example of Sensing Measurement Con

105、figurationIn the sensing mode of UE transmitting-UE receiving,both the transmitter and the receiverreceive the sensing measurement configuration from the base station,as shown in Figure 2-6.Figure 2-6 Example of Sensing Measurement ConfigurationAlternatively,the UE receiving the sensing signal obtai

106、ns the sensing measurement configurationfrom the UE transmitting the sensing signal,as shown in Figure 2-7.16/90Figure 2-7 Example of Sensing Measurement ConfigurationFor the sensing mode of base station transmitting-base station receiving,the transmitting basestation may need to negotiate with the

107、receiving base station to determine the sensingmeasurement configuration,as both have the authority to alocate resources.An independentdecision by one base station regarding sensing resources might affect the resource usage of theother base station.Therefore,for the sensing mode of base station tran

108、smitting-base stationreceiving,the measurement configuration process may include the following situations:The base stations directly negotiate to determine the sensing measurement configuration,as shownin Figure 2-8.Figure 2-8 Example of Sensing Measurement ConfigurationThe transmitting and receivin

109、g base stations negotiate with each other through AMF/SF andfinalize the sensing measurement configuration,as shown in Figure 2-9.Figure 2-9 Example of Sensing Measurement Configurationd)Sensing procedure in inactive modeFor bistatic sensing services involving a UE,there might be minimal demand for

110、communicationdata transmission.If the UE can support sensing operations in inactive mode,it would be17/90significantly beneficial in reducing power consumption.For example,utilizing downlink sensingsignals from the base station for respiration and intrusion detection,and employing uplink sensingsign

111、als from the UE for environmental reconstruction,all require the UE to perform sensingoperations while in inactive mode.If the UE supports the sensing in inactive mode based on downlink signals,the correspondingsensing procedure is shown in Figure 2-10.Figure 2-10 UE Sensing Procedure in Inactive Mo

112、de Based on Downlink SignalsOnce the sensing service is triggered,the network can configure the UE to perform sensingoperations in inactive mode based on its reporting capabilities(e.g.,supporting sensing signalreception and processing).If there is no requirements for communication data transmission

113、,the UE remains in inactive mode.The base station transmits sensing signals according to the configuration,while the UE receivesand processes the data as configured.During sensing operations,the UE must also handlecommunication-related tasks,such as synchronization,paging reception,and measurements.

114、When reporting sensing data is necessary,one method involves the UE entering connected modeto report the data according to network configurations,as shown in Figure 2-10(a).This methodposes no specific requirements regarding the amount of sensing data or reporting methods(e.g.,periodic or event-trig

115、gered reporting),but the UE needs to enter connected mode through randomaccess,which entails additional signaling and may introduce significant time delays.Alternatively,if the amount of sensing data is relatively small,the UE can report it through Small DataTransmission(SDT,Rel-17)in inactive mode,

116、as illustrated in Figure 2-10(b).If a communication service is required during this procedure,the UE needs to enter connectedmode while continuing the sensing operation.At this time,the sensing measurement configurationand data reporting must align with the settings defined in connected mode.For sen

117、sing operations involving uplink signals in the mode of UE transmitting-base stationreceiving,the UEs procedure in inactive mode also needs to be taken into account.Thecorresponding sensing procedure is shown in Figure 2-11.18/90Figure 2-11 UE Sensing Procedure in Inactive mode Based on Uplink Signa

118、lsSimilar to the sensing based on downlink signals,once the sensing service is triggered,thenetwork can configure the UE to perform sensing operations in inactive mode based on itsreporting capabilities(e.g.,supporting sensing signal transmission).If there is no requirements for communication data t

119、ransmission,the UE remains in inactive mode.The UE transmits uplink sensing signals according to the specified configuration,as illustrated inFigure 2-11(a).During this procedure,the UE must coordinate the transmission of sensing signalswith communication-related operations,while the base station ha

120、ndles signal reception andprocessing.Alternatively,as depicted in Figure 2-11(b),the UE can transmit uplink sensingsignals through Small Data Transmission(SDT)in inactive mode.When the UE is in idle mode,the radio access network lacks information about its status,raisingthe question of how to effect

121、ively trigger the UE to participate in sensing activities,which requiresfurther investigation.2.3 Measurement Quantities Definition and Sensing InformationTransmission2.3.1 Definition of Sensing MeasurementQuantitiesIn the 6G era,sensing technology is emerging as a key component in the realm of info

122、rmation andcommunication,particularly in bistatic sensing scenarios.Bistatic sensing differs from monostaticsensing in how basic measurement quantitiessuch as time delay,Doppler,and anglearecalculated and defined during target detection.These differences arise from the complex pathpropagation charac

123、teristics and interference environments among multi-static nodes,makingtraditional measurement quantities inadequate for fulfilling complex sensing requirements.Consequently,in view of diverse sensing services,it is essential to define measurement quantitiesthat address specific sensing needs to sup

124、port the efficient implementation of collaborativesensing.Sensing information may encompass multiple hierarchical levels.An example of this hierarchicaldivision is presented below:-Sensing results:range,velocity,location information,trajectory,and vital signs such asrespiration and heartbeat.Additio

125、nally,it encompasses gesture/action recognition,vehicle19/90inspection data,smart intersection details,dynamic maps,imaging results,weather conditions,airquality,shape,material composition,and more.-Preliminary sensing data:delay,Doppler,angle,and signal intensity(power).It may also consistof spectr

126、al information,such as delay spread spectrum,Doppler spectrum,micro-Dopplerspectrum,angle spectrum,or combinations of these,such as time delay-Doppler spectrum,timedelay-angle spectrum,or time delay-Doppler-angle spectrum.-Raw sensing data:received signal or original channel information(such as comp

127、lex-valuedresult,amplitude and/or phase of received signal or channel response,I/Q channels and theirderived results).3GPP TS 38.215 defines commonly used measurement quality indicators in communicationsystems,including Reference Signal Received Power(RSRP),Reference Signal Received Quality(RSRQ),an

128、d SINR10.Taking SINR as an example,it is defined as the ratio of signal power on aspecific resource element to the noise and interference power.When ignoring interference,SINRcan be substituted with SNR,which is calculated as the signal power on a specific resourceelement divided by the noise power.

129、In communication systems,SNR is a crucial indicator ofsignal transmission quality and closely correlates with communication performance.Generally,ahigher SNR indicates better signal transmission quality and improved bit error rate(BER)performance.For a communication system,the signal power in SNR(or

130、 SINR)encompasses thetotal received signal power propagated through LOS and multipath channels associated withvarious reflectors or scatterers.For a sensing system,the signal power in SNR(or SINR)specifically refers to the received power propagated through multipath channels associated withsensing t

131、argets,i.e.,the power of signals reflected by the sensing target.Different sensing servicesentail variations in signal power calculations.For instance,in detecting vehicles,pedestrians,orUAVs,the signal power refers to the received signal reflected by dynamic targets,while signalsreflected by static

132、 targets are considered static clutter.In contrast,for static environmentreconstruction or obstacle detection,the signal power includes the received signal reflected bystatic targets.For sensing SNR(or SINR),it is essential first to define the signal power(numerator)associatedwith the sensing target

133、.For example,when reporting point cloud information,SNR or SINR canbe expressed in terms of scattering points.The numerator is the power of each scattering point,and for the same target,the signal power of each scattering point may be different.Secondly,noise,interference and clutter(the denominator

134、)should be defined.This includes thermal noise,interference(communication or other sensing signals)and dynamic clutter,which cannot becanceled through sensing algorithms and can be assumed to be uniformly distributed within acertain delay,Doppler or angle domain(such as within the observable range o

135、f the sensingsignals).Since SNR and SINR are mainly used to characterize the signal quality of statisticalsignificance,noise and interference can be calculated by averaging the power in areas devoid of20/90targets,clutter,or interference,as the latter two can be canceled.It should be noted that the

136、LOSpath(the direct path between the transmitting node and the receiving node for bistatic sensing)andstatic clutter can usually be canceled through sensing processing,which do not significantly affectthe sensing results and may not be included in the denominator.Therefore,when estimating theSINR,it

137、is necessary to implement some cancellation algorithms for LOS path,interference andclutter first,and then calculate the average noise and interference power.Figure 2-12 presents the evaluation results of sensing performance along with the correspondingstatistical results of communication SNR and se

138、nsing SNR.In this simulation,the TRP-UEbistatic sensing mode is employed.Due to varying UE locations,the observed RCS of sensingtargets and the path loss differ,resulting in diverse sensing performances across UEs.The resultsindicate that sensing SNR more effectively reflects sensing performance qua

139、lity compared totraditional communication SNR.Thus,it is recommended to incorporate new signal qualityindicators,such as sensing SNR and sensing SINR,in the design of sensing systems.Theseindicators can provide a more accurate assessment of sensing signal transmission quality,helpevaluate the reliab

140、ility of sensing measurement results,and serve as reference indicators forequipment selection,sensing beam management,and resource configuration adjustments.Figure 2-12 Statistics on Sensing Accuracy and SNR:Positioning Accuracy(left),CommunicationSNR(middle),Sensing SNR(right)More broadly,the conce

141、pt of SINR-P(Signal-to-Interference-plus-Noise Ratio per Path)can beintroduced as a key performance indicator in a sensing system.SINR-P represents the ratio ofsignal to interference and noise for any given path,providing a comprehensive measure of signalquality for each echo path by taking into acc

142、ount both actual path loss and interference conditions.In sensing tasks,SINR-P can be used to select collaborative nodes to improve sensing accuracy.For example,in the process of collaborative sensing,node A detects a target through a wide beam,and the echo signal reflected by the target is received

143、 by its neighboring nodes B,C,D,etc.Adjacent nodes may report the SINR-P of the first path farther than the direct path or the first paththat is farther than the direct path and has a non-zero velocity for comparison and analysis by nodeA.Finally,node A selects the cooperative node with the highest

144、SINR-P(such as node D)toimprove the robustness and accuracy of target sensing.Figure 2-13 illustrates this process,emphasizing the importance of inter-nodal collaboration and the value of SINR-P in a realscenario.21/90Figure 2-13 Schematic Diagram of Cooperative Node Selection2.3.2 Protocol Stack of

145、 Sensing Information TransmissionThe transmission of sensing information involves a protocol stack that varies based on thetransmitting/receiving nodes or the volume of transmitted content.Options include a controlplane-based protocol stack,a user plane-based protocol stack,or a data plane-based pro

146、tocol stack.To simplify design,the standard should clearly define the protocol stack used for sensinginformation transmission,eliminating the need for configuration or negotiation.The following isan example of the first method for transmitting sensing information via the control plane,in whichthe po

147、tential protocol stack designs between the UE and SF,as well as between the base stationand SF,are provided.For the transmission of sensing information between the UE and SF,the potential protocol stack isshown in Figure 2-14,where XX represents the version number of 6G,7G,etc.It should be notedthat

148、 the figure includes a protocol stack between the UE and the base station.For the transmissionof sensing information between the UE and the base station,the first method can utilize RadioResource Control(RRC)signaling based on the control plane or transmit it through the datachannel.Figure 2-14 Exam

149、ple of Sensing Information Transmission Protocol Stack Between UE and SF22/90The potential protocol stack design for the transmission of sensing information between the basestation and the SF is illustrated in Figure 2-15.Figure 2-15 Example of Sensing Information Transmission Protocol Stack Between

150、 the BaseStation and SFThe potential protocol stack design for the transmission ofdesign of the protocol stack fortransmitting sensing information between base stations is shown in Figure 2-16.Figure 2-16 Example of Sensing Information Transmission Protocol Stack Between Base Stations2.3.3 Reporting

151、 of Measurement ResultsISAC network is a key technology for 6G.Target detection,high-accuracy positioning,environment reconstruction,and imaging supported by communication or sensing signals can bewidely applied in indoor positioning,UAVs,intelligent vehicles,and the Internet of Things,facilitating

152、the interconnection of various devices.In existing protocols,there are somepositioning methods,namely the location management function(LMF)determines the location ofa target based on measurement quantities such as RSRP,delay(difference),and angle.However,to accommodate a broader range of sensing ser

153、vices,additional measurement quantities arerequired,which increases demands on the reporting order of these quantities.Whether for23/90independent or collaborative sensing,numerous measurement quantities can be obtained duringtarget detection,making efficient reporting a significant concern.In the m

154、obile network,SF needs to perform sensing based on network requirements.To addressissues of frequent interactions and low resource utilization in high-accuracy collaborative sensing,it is considered to design a new sensing measurement quantity reporting mechanism.Thismechanism establishes a priority

155、 sorting scheme for measurement quantities and a datacompression method,ensuring clear understanding between the transmitter and receiver.It retainsessential interactive content and enhances resource utilization.In the sensing task,multiple paths and measurement quantities need to be reported.The SF

156、configures the reporting order based on the capabilities of the sensing nodes,including the priorityof both the configured paths and measurement quantities.This priority order can be definedthrough network configuration,enabling the sensing node to report multiple measurementquantities across variou

157、s paths.Path 1:Measurement quantitiy 1,2,.,P.Firstly,the sorting issue of multiple paths with multiple measurement quantities is discussed,andtwo sorting methods for measurements are proposed.1)Sorting method 1:Path first,then measurement quantitiy,that is,all measurement quantities aresorted in seq

158、uence according to each path,as shown in Figure 2-17,where resource inforepresents some fixed quantities,and Am,Bm,and Cmrepresent different measurement quantitiesof the mthpath respectivelyPath M:Measurement quantitiy 1,2,.,P.Figure 2-17 Sorting Method 12)Sorting method 2:Measurement quantitiy firs

159、t and then path,that is,all paths are sorted insequence according to each measurement quantitiy,as shown in Figure 2-18.Am,Bm,.and Xmrepresent different measurement quantities of the mthpath respectively.Figure 2-18 Sorting Method 2In actual network environments,resource constraints may hinder sensi

160、ng reporting.Therefore,it isessential to consider the priority of paths and measurement quantities.Based on prioritization,thereporting content is categorized into Sensing-1 and Sensing-2.Sensing-1 is fixed and includesessential resource information and high-priority measurement quantities.If it is

161、not fully utilized,the corresponding bits are set to 0.Sensing-2 comprises additional measurement quantities withlower priority.24/901)Put all the measurement quantities of the first m paths with higher priority in front,that is,Sensing-1;2)Put the first p measurement quantities with a higher priori

162、ty of M paths in front,i.e.Sensing-1;3)Put the first m paths with higher priority and the first p measurement quantities with higherpriority in front,that is,Sensing-1;In addition,a data compression scheme for reporting sensing measurement quantities is proposed,as shown in Figure 2-19.For the first

163、 path,all measurement quantities will be reported.Forsubsequent paths,the measurement quantities are subtracted from those of the previous path.If thedifferences in measurement quantities needed at present fall within a specified range,only thedifferences compared to the previous path need to be rep

164、orted.Figure 2-19 Schematic Diagram of Data CompressionSensing data can be transmitted efficiently and accurately in high-accuracy collaborative sensingby configuring reporting rules for measurement quantities.This approach ensures that both partieshave a clear understanding of the measurement quant

165、itiy combinations while prioritizing thetransmission of more critical sensing data.25/903Bistatic Sensing Key Technologies3.1 Signal Design3.1.1 Sequence DesignThe peak-to-average power ratio(PAPR)of the orthogonal frequency division multiplexing(OFDM)waveform in existing communication systems is hi

166、gh.And power amplifiers mustoperate in a backed-off state to remain within their linear region,leading to reduced transmissionpower.This reduction directly affects the strength of the sensing echo signal,making it harder todetect.To enhance performance,an ISAC system based on the OFDM waveform shoul

167、d ideallyutilize low-PAPR sequences for sensing.This approach can reduce power back-off,increasetransmission power,and improve echo signal strength.The basic sequences used for the reference signal in current communication systems include theZadoff-Chu(ZC)sequence,maximal length linear feedback shif

168、t register(m)sequence,and theGold sequence.Among these,the ZC sequence exhibits a constant modulus characteristic,and itsdiscrete Fourier transform remains a ZC sequence,yielding a lower PAPR.The ZC sequence alsopossesses excellent self-correlation and cross-correlation properties,with correlation v

169、aluesbetween the ZC sequence and its cyclic shifts being zero,and minimal cyclic cross-correlationamong different root sequences.Additionally,the ZC sequence demonstrates high Dopplertolerance,providing robust performance against frequency offsets and excelling in detectingechoes from hig-speed targ

170、ets.For instance,the basic sequence employed for uplink soundingreference signals(SRS)in the 5G NR communication system is the ZC sequence.Another commonly used sequence is the Gold sequence,a pseudo-random sequence that serves asa key reference signal sequence in the 5G NR communication system.The

171、Gold sequence isgenerated by performing modulo-2 addition on two m-sequences,followed by 2-bit QPSKmodulation to create a complex signal sequence.The Gold sequence forms the basis for referencesignals such as the CSI-RS(Channel State Information-Reference Signal),DMRS(DemodulationReference Signal),a

172、nd PRS(Positioning Reference Signal).It exhibits good self-correlation andcross-correlation properties,as well as effective resistance to frequency offsets,supporting a largenumber of sequences,though its PAPR is relatively high.Both the Gold sequence and the ZCsequence can serve as candidate sequen

173、ces for sensing signals.In communication systems,OFDM symbols are typically generated and mapped individuallywithout much consideration of the correlation between sequences of different symbols.As a result,this approach often leads to insufficient sequence randomness in the time domain.In sensingsys

174、tems that utilize two-dimensional joint processing in both the time and frequency domains,thecorrelation characteristics of different symbol sequences in the time domain are suboptimal.Toaddress this issue,scrambling or randomizing different symbol sequences in the time domain canhelp optimize their

175、 characteristics and reduce cross-correlation peaks11.Assume that each26/90sensing coherent processing interval contains M OFDM symbols,each carrying a reference signalsequence of length N,represented as=0,1,1,0 1.Byapplying randomly generated coefficients ,0 1,the reference signal sequencesacross d

176、ifferent symbols can be randomized ,=,0 1,0 1.This new signal maintains the sequence characteristics of the reference signal in thefrequency domain while improving its characteristics in the time domain due to the overallrandomization.The generation methods for the CSI-RS Gold sequence and the SRS Z

177、Csequence(with group hopping enabled)in 5G NR are used as a baseline,and the cross-correlationcharacteristics of these sequences after randomization are compared,as depicted in Figure 3-1.The results demonstrate a significant improvement in the cross-correlation characteristics of theoptimized signa

178、l in the time domain.Figure 3-1 Comparison of Cross-correlation Performance Before and After Sequence OptimizationThe sensing performance of the sequences before and after optimization is evaluated throughsystem simulation,and interference from neighboring cells is modeled in the simulation.The ZCse

179、quence was further considered under three different configurations:without group hopping orsequence hopping,with group hopping,and with sequence hopping.The simulation results areshown in Figure 3-2，indicating that both the Gold sequence and ZC sequence exhibitsignificantly improved target positioni

180、ng performance compared to the existing sequencegeneration methods used in 5G NR,following optimization of the sequences in the time domain.Figure 3-2 Comparison of Positioning Performance Before and After Sequence OptimizationIn contrast to communication systems,the echo signals reflected from targ

181、et objects are typicallyvery weak when received by sensing nodes.Power imbalances in the time domain can easily27/90distort these weak signals,adversely affecting sensing performance.Therefore,it is preferable toutilize sequences with constant modulus and balanced time-domain power characteristics f

182、or thesensing reference signals.Although the ZC sequence inherently possesses a constant moduluscharacteristic,the actual reference signal sequences generated from the ZC sequence may notmaintain this property.For instance,in the case of the SRS,the practical SRS sequence lengthwill be greater than

183、the ZC root sequence length,and there is a phase cyclic shift item,so that the actual SRS sequence no longer exhibits constant modulus characteristics,asshown in Figure 3-3.Figure 3-3 Schematic Diagram of SRS Sequence Without Constant Modulus CharacteristicsOne approach to maintaining the constant m

184、odulus characteristics of sequences is to apply aDiscrete Fourier Transform(DFT)prior to their use.This allows the practical reference signalsequence to regain constant modulus characteristics,as illustrated in Figure 3-4.Figure 3-4 Schematic Diagram of Sequence with Constant Modulus Characteristics

185、 After DFTTransformationIn the current 5G communication system,ZC sequences are generated using the formula =exp j+1,where=0,1,2,1.For SRS,the calculation method of rootsequence index is =?+1 2+12?,?=+131.Here,0,1,29 isthe group index and 0,1 denotes the root sequence index within the group.Although

186、 theSRS generation supports group hopping and sequence hopping mechanisms,the maximum28/90number of group indexes and sequence indexes within a group is limited,restricting the amount ofavailable ZC sequences.To enable multi-target and multi-equipment sensing in future systems,a large pool of availa

187、bleZC sequences is essential.To achieve high range resolution,sensing systems typically utilizewider bandwidth resources and correspondingly longer sequences.Therefore,if the ZC sequenceis employed as the foundational sequence for generating sensing signals,the number of availableZC sequences can be

188、 expanded by revising the designs for group hopping or sequence hopping tosupportmoregroupindexesandsequenceindexes.Thisenhancementfacilitatestherandomization of interference among cells or sensing equipment,thereby improving overallsensing performance.3.1.2 Resource Mapping SchemesThe sensing signa

189、l resource mapping schemes need to be designed to meet performancerequirements such as sensing resolution and sensing measurement range.Greater signal bandwidthand longer signal duration enhance both range and velocity(Doppler)resolution,while increasingfrequency domain density and time domain densi

190、ty expands the unambiguous distance andvelocity ranges.Traditional radar sensing typically employs signals with continuous resources,large time widths,and substantial bandwidths.However,in ISAC systems,the impact oncommunication rates must also be taken into account,leading to limitations on the ove

191、rhead forsensing resources.Consequently,it is necessary to design signal resource mapping patterns thatalign with the specific requirements for the aforementioned sensing performance in variousservices.Uniform sensing signals offer relatively straightforward signal configuration and processing,provi

192、ding good sensing performance.However,in ISAC scenarios,uniform sensing signals faceseveral challenges12:High Time-Frequency Resource Overhead:To satisfy requirements for delay resolution,Doppler resolution,and maximum unambiguous measurement range,uniform sensing signalsconsume a substantial number

193、 of subcarriers and OFDM symbols.The resource overheadissues are exacerbated in multi-port sensing scenarios.Lack of Flexibility:Achieving high-resolution sensing performance requires a largetime-frequency resource span,and uniformly sampled sensing signals may occupy periodicsignalresources.Inenvir

194、onmentswithdiversecommunicationservices(includingultra-reliable low-latency communication)alongside sensing services,it becomes challengingto ensure that periodic signal resources can always be allocated to specific sensing signals.Inaddition,it is difficult for a uniform sensing signal to make full

195、 use of variouscommunication reference signals.In sensing applications,the targets are typically sparse in both the delay and Doppler domains,allowing recovery signals to be reconstructed from sampling points below the Nyquist sampling29/90rate.This enables the use of non-uniform signals for sensing

196、 services,effectively addressing thechallenges associated with uniform sensing signals.Related literature has extensively studied andapplied DOA estimation and beamforming techniques based on non-uniform antenna arrays.Consequently,the design of non-uniform signals and corresponding signal processin

197、g methodshas matured alongside research on these antenna arrays.In ISAC applications,non-uniform signaldesign can be extented from the spatial domain to the time-frequency domain for allocatingtime-frequency resources of the sensing signals.This approach reduces resource overhead whileenhancing the

198、flexibility of resource allocation.The design methods for non-uniform sensingsignals can be categorized into two main types:Non-Uniform Sensing Signal Design Based on Compressed Sensing:If the sensing signal issparse in its original or transformed domain,it can be reconstructed from samples at a rat

199、esignificantly below the Nyquist rate13.The key to this design approach lies in constructingthe sparse and observation matrices,which must satisfy the Restricted Isometry Property(RIP)as defined in compressed sensing theory14.In ISAC systems adopting OFDMwaveforms,the most typical sparse bases are t

200、he DFT sparse basis and the IDFT sparse basis.Specifically,for frequency domain signals converted to delay domain signals,the DFT sparsebasis applies.Conversely,for time domain signals converted to Doppler domain signals,theIDFT sparse basis is used.In addition,when allocating time-frequency resourc

201、es for anOFDM signal,each subcarrier or OFDM symbol can be in one of two states:allocated to thesensing signal or not.This binary state can be represented by a Bernoulli matrix.Consequently,the Bernoulli matrix can serve as an observation matrix,meeting the RIPcondition with either the DFT or IDFT s

202、parse basis.By applying these methods,the design ofnon-uniform sensing signal circuits can be simplified.At the receiver,an optimal solutioncan be obtained by iterative search using a greedy algorithm such as the orthogonal matchingpursuit algorithm(OMP).Non-Uniform Sensing Signal Design Based on Di

203、fferential Cooperative Array:Virtual arrayelements are constructed from the frequency or time differences between physical arrayelements,enabling the construction of a larger array with fewer physical elements15.Typical construction methods for virtual arrays include nested arrays and coprime arrays

204、,asillustrated in Figure 3-5 below,which shows both the physical and corresponding virtualarrays.At the receiver,the signals from the uniform virtual array are constructed byperforming covariance operations on the signals from the non-uniform physical array.Oncethe virtual array is established,furth

205、er processing can be conducted using methods such asDFT/IDFT or MUSIC.It should be noted that constructing the virtual array makes each pathbecome a coherent signal.Consequently,decoherence processing must be performed beforeapplying subspace algorithms like MUSIC.The non-uniform sensing signal desi

206、gn method isapplicable for one-dimensional sensing signal design,while independent two-dimensionaldifferential cooperative array design methods require further study.30/90Figure3-5 Schematic Diagram of Differential Collaborative ArraysFigure 3-6 compares the delay-Doppler spectrum obtained from prot

207、otypes using non-uniform anduniform sensing signals.It is clear that while the non-uniform sensing signal significantly reducestime-frequency resource overhead,it also results in a reduction in the sensing SNR.Therefore,practical applications must strike a balance between resource overhead and the s

208、ensing SNR.(a)Uniform Signal(b)Compressed Sensing Method(c)Nested Array MethodFigure 3-6 Measured Delay-Doppler Spectrum of Non-Uniform Sensing SignalSpecifically,a two-step non-uniform sensing signal design method can be utilized.First,auniform sensing signal is designed,followed by non-uniform sam

209、pling based on the uniformsignal to obtain the non-uniform sensing signal.Specifically,non-uniform sampling is conductedon the time-frequency resources occupied by the uniform signals,selecting a subset of subcarriersor OFDM symbols to carry the sensing signal.This process enables the design of the

210、non-uniformsensing signal,as illustrated in Figure 3-7.In comparison to the uniform signal,the non-uniformsignal reduces the number of signal resource elements required for the sensing signal andstrategically avoids certain subcarriers or OFDM symbols to prevent interference with othersignals.Figure

211、 3-7 Two-Step Non-Uniform Sensing Signal Design Method31/90The first step of the two-step non-uniform sensing signal design method addresses therequirements for delay and Doppler resolution,as well as the maximum unambiguousmeasurement range of the sensing signal,allowing for greater flexibility in

212、the subsequentnon-uniform sampling design.Additionally,the uniform signal configuration in this first step canbe easily implemented using existing resource allocation methods.Considering that there is a trade-off between the maximum unambiguous range and the sensingaccuracy,i.e.the ranging error is

213、linearly inversely proportional to(the interval betweensubcarriers occupied by the sensing signal),the maximum ranging range is also linearlyinversely proportional to .As illustrated in Figure 3-8,frequency domain resources areflexibly allocated based on sensing requirements,and the reference signal

214、 mapping patterns aredynamically adjusted.The receiver selects sensing algorithms based on different configurations.This approach effectively addresses the trade-off between ranging accuracy and maximum range,while reducing overhead,improving time-frequency resource utilization,and meeting the sensi

215、ngand ranging requirements of different scenarios.Figure 3-8 Non-Uniform Pilot-Frequency Adaptive Mapping SchemesTo obtain accurate terminal position information,velocity or other sensing information,thesensing reference signal often needs to be configured with a significant amount of time-frequency

216、resources.To reduce the sensing resource overhead,comb mapping is a potential solution.Asshown in Figure 3-9,comb mapping involves distributing sensing reference signals at regularintervals in both the time and frequency domains.When the sensing bandwidth and frame lengthare fixed,the comb mapping s

217、cheme can reduce the resource overhead without sacrificing sensingaccuracy and resolution.Figure 3-9 Time-Frequency Uniform and Equally Spaced Reference Signal Resource Mapping32/90The comb mapping scheme will introduce estimation ambiguity issues.For example,based on thearrangement of equally space

218、d sensing reference signal,the maximum ambiguous range can bederived as=2adjacent reference signals in the frequency domain.The maximum ambiguous velocity is given by=where is the subcarrier spacing and represents resource elements(RE)interval of two2where represents the symbol interval and represen

219、ts the symbol interval of adjacent reference signals in the time domain.It can be observed that a smaller frequency domain interval of sensing signals results in a larger unambiguous distance range,while a shorter time domain interval leads to a larger unambiguous velocity range.Similarly,to address

220、 issues of ambiguous sensing,a non-uniform distribution of the sensing signals can be implemented,employing a time-frequency sparse sensing reference signal design.For a given sensing bandwidth and sensing frame length,the non-uniform arrangement of sensing signal units is adopted for sparse sensing

221、 signals to avoid ambiguous estimation while maintaining sensing accuracy and resolution.At the same time,with a fixed number of sensing reference signal units,non-uniform sensing signals can achieve greater accuracy and resolution.Typical non-uniform sensing signal patterns include coprime-based sp

222、arse patterns and nesting-based patterns:Coprime-Based Sparse Sensing Signal Pattern:This pattern consists of two uniformlyequally spaced sensing signal patterns,one of which contains 2M sensing signal units,andthe interval between adjacent sensing signal units is Nd,and another of which contains Ns

223、ensing signal units,and the interval between adjacent sensing signal units is Md.Thefollowing schemes can be adopted for the pattern design:=+,=0,1,2,1 +,=0,1,2,2 1where represents the set of all sensing signal units,represents the minimum intervalbetween two adjacent sensing reference signal units.

224、,zero-padding of the input signal is required.A two-dimensional periodogram is applied for sensing parameter calculation,with the input signalbeing the sensing information detection matrix.The two-dimensional periodogram is obtainedthrough Inverse Fast Fourier Transform(IFFT)and FFT as follows18.,=1

225、=01=01,2?2?2The sinusoidal components related to the reflection path delay and Doppler of the sensing target inthe sensing information detection matrix will produce corresponding peaks in,.If41/90?,?corresponds to one such peaks,the estimated delay and Doppler can be calculated asfollows?=?,?=?where

226、 represents the speed of light,is the OFDM subcarrier spacing,and is the carrierfrequency.(2)Capon algorithmThe Capon algorithm is commonly used for target angle estimation.Capon is a DOA estimationalgorithm based on array signal processing.It calculates the covariance matrix of the input signaland

227、the spatially spectrum function related to the array direction vector and performs a peaksearch to obtain angle estimation values.Typically,the input signal is derived from therange/Doppler pulse compression results of different antennas,such as the periodogram result?,?.The optimization problem tha

228、t the Capon algorithm addresses can be expressed as|()|2=min where()is the input signal,is the array element weight vector,and is the covariancematrix of the input signal.Further calculations yield the spatially spectrum function of Capon asPCapon=1aHkR1a kwhere is the steering vector of the desired

229、 signal,representing the signal direction vectorrelated to angle.The spectrum peak is identified based on the calculations of the spatially spectrum function,andthe index of the peak indicates the targets angular information.(3)MUSIC algorithmMUSIC algorithm is a subspace decomposition based algorit

230、hm that utilizes the orthogonalitybetween the signal subspace and noise subspace to construct a spatially spectrum function,followed by spectral peak searching for parameter estimation.Based on the sensing informationdetection matrix,the delay and Doppler auto-covariance matrices are estimated as fo

231、llows19.R=1MFFH NN,Rfd=1NFHF MMEigenvalue decomposition is performed on and.The eigenvectors corresponding to the maximum eigenvalues constitute a signal subspace,while the eigenvectors corresponding to the remaining eigenvalues constitute a noise subspace,expressed as .The steering vector correspon

232、ding to the targets angular frequency =is orthogonalto the noise subspace,that is,it is located in the null space of the noise subspace.Thefollowing formula can be solved to obtain?,and?,.sHz VVHs z=0According to?,and?,the angular frequencies?,=arg?,and?,=arg?,associated with the reflection path del

233、ay and Doppler of the sensing target are obtained,allowingfor the estimation of the targets delay and Doppler.42/90?=?2f,f?d=?fd2TO(4)ESPRIT algorithmThe ESPRIT algorithm is a high-resolution spatial domain spectrum estimation method based onrotation invariance,also classified as a classical subspac

234、e estimation algorithm.Based on thesensing information detection matrix ,the delay and Doppler auto-covariance matrices areestimated asR=1MFFH NN,Rfd=1NFHF MMTakingas an example,perform eigenvalue decomposition to obtain the sorted eigenvalues=diag 0,1,1.The eigenvectors corresponding to the maximum

235、 eigenvalues in form a signal subspace,and the eigenvectors corresponding to the remaining eigenvalues form a noise subspace,as followsU=SV,S NP,V N NPBy extracting the first 1 and last 1 rows of the signal subspace,we obtainS1=IN1O S,S2=OIN1Swhere 1is an(1)(1)identity matrix,is a vector of length N

236、-1 with allelements equal to 0.Based on 1and 2,further calculations yield=S1HS11S1HS2where the eigenvalues of the matrix can be used to estimate angular frequencies associatedwith range or velocity.and are obtained by calculating the delay auto-covariance matrixand Doppler auto-covariance matrix res

237、pectively,and then the estimation results for thedelay and Doppler frequency of the sensing targets reflection path can be calculated according tothe maximum eigenvalues?,and?,in and.?i=arg?,i2f,f?d,i=arg?fd,i2TO,i=0,1,P 13.2.2 Position Information Calculation for Bistatic SensingThe angles,ranges a

238、nd positions of the target and the transceiver nodes under bistatic geomotryare shown in Figure 3-21.43/90 xyzTargetAOATZOATzxyRAOALZOALTxRxdTx-TargetdTarget-RxFigure 3-21 Spatial Relationship Between Target and Transmitting/Receiving Nodes in BistaticSensing ModeThe receiver measures the azimuth an

239、gle of arrival and zenith angle of arrival ofthe sensing targets reflected path,the azimuth angle of arrival and zenith angle of arrivalof the LOS path,and the delay difference between the sensing targets reflected pathand the LOS path.The receiver calculates the bistatic range sum based on the dela

240、y differencebetween the targets reflection path and the LOS path,alongside the distance between thetransmitter and receiver.dTxTarget+dTargetRx=c +dTxRxwhere is the speed of light and denotes the distance between the transmitter andreceiver,known as the bistatic baseline distance.Next,using the DOA

241、information for both the LOS path and the sensing targets reflected path,the angle between the two incoming waves in the bistatic plane is calculated.R=cos1sin ZoATsin ZoALcos AoAL AoAT+cos ZoATcos ZoALThe distance between the target and the receiver can then be determined based on the anglebetween

242、the two incoming waves and the bistatic distance.dTargetRx=dTxTarget+dTargetRx2 dTxRx22 dTxTarget+dTargetRx dTxRx cos RFinally,the position coordinates,of the sensing target relativeto the receiver can be calculated using both the targets distance from the receiver and the angularinformation of the

243、target path,wherein=44/90For bistatic sensing,the DOA information measured by the receiver is typically relative to its localcoordinate system.To calculate the targets position in the global coordinate system,it isnecessary to transform the DOA information from the local coordinate system to the glo

244、balcoordinate system,and the transformation relationship between the two coordinate systems isrequired.For bistatic sensing,position information of the target can be obtained not only through themeasurement of DOA and delay,but also by measuring the departure angle and delay.Bycalculating the distan

245、ce from the target to the transmitter using these measurements,the targetsposition relative to the transmitter can then be determined based on the derived distance anddeparture angle.In bistatic sensing mode,the antenna configuration of the transmitting equipmentis typically unknown to the receiver.

246、If angle of departure measurements are needed by thereceiver,the transmitting antenna configuration must be informed,or the angular information canbe indirectly obtained through receiver-side precoding vector search or beam measurement,usingthe association between precoding vectors or beams and angl

247、es.In addition,bistatic sensing provides different observation angles compared to monostatic sensing.The bistatic triangle can be determined only by measuring the departure angle and DOA,and theposition information of the target can be calculated in combination with the bistatic baselinedistance.In

248、multi-static sensing modes,targets position information can be obtained bymeasuring either delay or angle.However,it is crucial to ensure that the measurement information,such as angles or delays,is correctly matched when multiple targets exist.3.2.3 NLOS-Based Sensing MethodNon-line-of-sight(NLOS)t

249、ransmission is a frequently occuring non-ideal factor that significantlyimpacts sensing accuracy.In the bistatic sensing process,line-of-sight(LOS)transmission occurswhen the sensing signal travels directly between the transmitter and the target,as well as betweenthe target and the receiver,without

250、reflection,refraction,or diffraction caused by surroundingscatterers,as illustrated in Figure 3-22(a).Conversely,NLOS transmission occurs when scatterers,such as buildings and trees,are present between the transmitter and the target,or between thetarget and the receiver.This results in the sensing s

251、ignal being reflected,refracted,or diffracted bythe scatterers,in addition to reflections from the target,leading to multi-hop reflection of thesignal.There are three approaches to handle the NLOS issue:identification,cancellation,andutilization.When the LOS path is blocked,the last approach is the

252、only one that can transformNLOS from a disadvantageous factor into an advantageous one,ensuring sensing accuracy underNLOS conditions.(1)The NLOS scenario in bistatic sensing is shown in Figure 3-22.The sensing transmission node(Base Station A)emits a sensing signal to detect a pre-identified target

253、 T,while the sensing echois received by the sensing reception node(Base Station B).The scatterer is denoted as S.Onlyreflections within two hops are considered.There exists a LOS path between A and T,but the45/90LOS path between T and B is obstructed by obstacles.Additionally,there is another scatte

254、rer(S)present in the environment,while the positions of both target T and scatterer S are unknown.Itcan be seen that there are 2 paths,one is the LOS path of transmitter-scatterer-receiver(A-S-B),and the other is the NLOS path of transmitter-target-scatterer-receiver(A-T-S-B).If thetraditional sensi

255、ng method(i.e.,AoA and delay estimation)is applied to the A-T-S-B path,it willyield the position of a false target,indicated as G in Figure 3-22.To determine the trueposition of the target,we must utilize the geometric relationship between the two paths.(1)Schematic Diagram(2)Geometrical Relationshi

256、p DiagramFigure 3-22 Collaborative Sensing Scenario under NLOSAs shown in Figure 3-22,the two paths share identical AoA but differ in AoD,which can be usedto distinguish the two paths.The geometric relationship between the two paths is as follows:(1)LOS path of transmitter-scatterer-receiver(A-S-B)s

257、in1=sin1+=11+1=0(2)LOS path of transmitter-target-scatterer-receiver(A-T-S-B)(+)sin2=(+)sin2(+)cos2+(+)cos2=02+2 2(2 2)=2+=2(3)If the LOS path of transmitter-target-receiver(A-T-B)exists,then(sin3)2+(0 cos3)2=2+=3In contrast,when a LOS path via the target exists,only AoA and delay are required to de

258、terminethe target position.However,when the LOS path is absent,the geometric relationships in theA-S-B path and the A-T-S-B path include delay and angular parameters,the scatterer pathdistance,and LOS and NLOS path distances that can be measured or estimated.Other distanceparameters must be inferred

259、 from these measured or estimated parameters using geometricrelationships.Therefore,to estimate the true position of the target using the NLOS path,asopposed to using the LOS path,it requires joint AoA-AoD estimation combined with delay46/90information.An NLOS utilization algorithm has been proposed

260、 for this purpose.The algorithmprocess is illustrated in Figure 3-23 and consists of the following five steps:1Joint AOA-AOD estimation:Traditional AOA/AOD angle measurement algorithms,such asMUSIC,can be extended to two-dimensional joint estimation,yielding two pairs ofAoA-AOD combinations correspo

261、nding to the signal paths.2A-S-B path distance estimation:Algorithms like 1-D DFT can be employed.3Signal reconstruction and interference cancellation for A-S-B path:Based on the estimateddelay and angular information,along with the path loss model,the A-S-B path signal can beinferred and removed fr

262、om the composite signal to obtain the A-T-S-B path signal.4A-T-S-B path distance estimation:Conduct a 1-D DFT on the A-T-S-B path signal to extractdelay and distance information.5Target position estimation:Leverage geometric relationships to determine the target position.Figure 3-23 Flow Chart of NL

263、OS Utilizing Algorithms in Collaborative SensingA simulation comparison and analysis were conducted for four cases in the 4.9 GHz frequencyband,with results illustrated in Figure 3-24,including several algorithms:1Proposed algorithm(with interference cancellation)2Proposed algorithm(without interfer

264、ence cancellation)3Traditional algorithm(with the algorithm for the traditional LOS path to process NLOS path)4Traditional algorithm(with LOS path as a benchmark)Figure 3-24 Position Estimation Results47/90It can be seen that the proposed method(red star)can estimate the true position of the target

265、and isclose to the estimated position in the A-T-B LOS case.The distance estimation result for theA-T-S-B path will be close to that of A-S-B path without interference cancellation(green star),suggesting the target is located between points A and S.Given the AOD direction aligns with A-T,the final e

266、stimated target position is near the intersection of these pathsessentially theoriginresulting in substantial error.Meanwhile,estimating the targets position(purple star)using the distance from the A-T-S-B path and AOA leads to the position of a false target.3.3 Non-Ideal Factors Cancellation3.3.1 H

267、ardware Non-Ideal Factors and ImpactsIn an ISAC system,sensing non-ideal factors are those that have minimal impact oncommunication but significantly impair sensing performance20.The suppression of thesenon-ideal factors is crucial for enhancing the performance of bistatic sensing and has been thefo

268、cus of extensive researches in recent years.In this section,we consider sensing non-ideal factorsprimarily stemming from time-frequency asynchronization between the transmitter and receiver ofsensing signals,including timing offset(TO),timing drift(TD),and carrier frequency offset(CFO).In addition,r

269、andom phase variations caused by equipment state switching also has asignificant impact on sensing performance.Timing offset:The transmitter and receiver generate clock signals using their own clocksources for timing synchronization.A discrepancy between these clock signals introducestiming offset,r

270、esulting in additional phase deviations in the frequency domain besides thosecaused by signal propagation delays.Consequently,timing offset leads to ambiguity in themeasured delay of sensing signals,leading to inaccuracies in range measurements.Timing drift:Due to variations in the sampling clock pe

271、riodicities of the transmitter andreceiver,as well as timing adjustments at the receiver,the timing offset varies over time.Thisresults in different timing offsets for different OFDM symbols,preventing coherentaccumulation across OFDM symbols21.Timing drift can be regarded as the temporalvariation o

272、f timing offset,in contrast to the previous discussion,which focused on the entiretiming offset for a specific OFDM symbol or coherent processing interval(CPI).Carrier frequency offset:This issue arise from two aspects.The first aspect is the localoscillator frequency deviation between the transmitt

273、er and receiver,as each use its ownfrequency source for generating signals.The second aspect is the relative movement betweenthe transmitter and receiver,which induces a Doppler frequency shift.The carrier frequencyoffset introduces ambiguity in Doppler measurements.Random phase:State switching(e.g.

274、,on-off switching,gain adjustments)in internalequipment modules(e.g.,power amplifiers)can lead to random phase variations.Similar totiming drift,random phase prevents the coherent accumulation across different OFDMsymbols,potentially affecting Doppler frequency or velocity measurements and sometimes

275、48/90leading to measurement failure.Random phase occurences are particularly prevalent inlow-cost equipment,such as budget terminals.Figure 3-25 illustrates the impact of non-ideal sensing factors on the sensing signal across a set ofmeasured data.In this dataset,both transmitting and receiving equi

276、pment were stationary,while amoving target presented in the environment.The received signal comprises the LOS path signal,the reflection path signal from the moving target,and signals from surrounding objects.Figure3-25(a)displays the delay spectrum impacted by these non-ideal factors,wherethe delay

277、 andamplitude discrepancies between OFDM symbols primarily result from timing drift.Furthermore,Figure 3-25(b)presents a delay-Doppler spectrum that is significantly dispersed,whichcomplicates the accurate detection of the sensing target.After eliminating the effects of timingdrift,the resulting del

278、ay-Doppler spectrum is shown in Figure 3-25(c).It reveals that the pathcarrying the majority of received power does not reside in the zero Doppler column,indicating aclear effect of carrier frequency offset.Additionally,the timing offset results in an overall shift inthe delay-Doppler spectrum in Fi

279、gure 3-25(c)along the delay axis.Figure 3-25 Influence of Sensing Non-Ideal Factors3.3.2 Hardware Non-Ideal Factor Cancellation Schemes(1)Cancellation-based ApproachesThe essence of sensing non-ideal factors lies in introducing additional phase errors.In theprevious researches on ISAC,mechanisms of

280、these non-ideal factors affecting sensing signalswere not fully revealed.The primary approach to suppress sensing non-ideal factors is eliminatingtheadditionalphaseerrorsintroducedbysensingnon-idealities,typicallythroughcancellation-based techniques.The following categories of cancellation methods a

281、re considered22:Cancellation in the Antenna Domain:It mainly includes two types:Cross AntennaCross-Correlation(CACC)and Cross Antenna Signal Ratio(CASR).As their names imply,these techniques perform cross-correlation and division operations on the CSI betweenantennas2324.The CACC method is often ref

282、erred to as the conjugate multiplicationmethod,while The CASR method is known as the CSI ratio method.These methods leveragethe fact that sensing non-ideal factors caused by clock asynchronization are identical across49/90different antennas.Consequently,the timing offset,timing drift and carrier fre

283、quency offsetexperienced by signals received by different antennas are the same,enabling the suppressionof non-ideal factors through signal processing in the antenna domain.In the case where thereference signal employs a constant modulus sequence(e.g.,ZC sequence),the outcomes ofconjugate multiplica

284、tion and division vary only by a constant coefficient.It is important tonote that for effective cancellation in the antenna domain,RF and clock generation modulesmust be shared among the antennas.In systems with a large number of RF channels,theremay exist multiple sets of RF and clock generation mo

285、dules,making inter-modulecancellation unavailable.This approach is particularly effective for extracting Dopplerinformation and has been successfully applied in sensing scenarios such as detecting vitalsigns(e.g.,respiration and heartbeat)with commendable performance.However,challengesarise when att

286、empting to obtain the delay and angular information of the sensing target tosubsequently determine its position.Cancellation in the Delay Domain:After transforming the channel state information to thedelay domain through an IDFT,the effects of non-ideal factors can be mitigated via phasecancellation

287、 among different paths25 26.This is because the non-ideal factors affectingdifferent paths remain consistent over a small time interval(e.g.,the duration of one OFDMsymbol).Typically,the phase contributions from all factors,excluding the Doppler frequencyof the moving target,can be extracted from th

288、e LOS path or the strongest static path.Thisextracted phase can then be applied to perform phase compensation on all other paths.Thisapproach is applicable in both single-antenna and multi-antenna scenarios.It effectivelycircumvents issues related to image frequency and nonlinearity associated with

289、antennadomain cancellation,making it an increasingly prefered solution.However,the method relieson the presence of a separable strong path,and its effectiveness highly depends on the powerof this reference path24.It is generally most effective in LOS conditions,where the LOSpath serves as the refere

290、nce for canceling non-ideal factors.Additionally,cancellation in thedelay domain typically necessitates higher delay resolution,which can be achieved byleveraging wideband signals in the millimeter-wave frequency range.Simulation resultsconfirm the effectiveness of the synchronization scheme based o

291、n the reference path,demonstrating excellent compensation performence for time and frequency asynchronization,and substantially improving range and velocity estimation accuracy,as illustrated in Figure3-26.It shows that the reference path synchronization scheme is feasible and efficient inpratical I

292、SAC(ISAC)scenarios.50/90Figure 3-26 Range and Velocity Estimation with and Without Time and Frequency SynchronizationCancellation in the Spatial Domain:Since static paths do not contain Doppler frequencycomponents from moving targets and only exhibit phase shifts due to non-ideal factors2728,it is p

293、ossible to identify a static path in the spatial domain to estimate the phase contributionsfrom all non-ideal factors for cancellation.This method is fundamentally similar to delaydomain cancellation,as both approaches utilize static paths to extract and eliminate non-idealfactors.However,this appro

294、ach requires a large degree of freedom in the spatial domain toeffectively distinguish the static path from other paths.Existing communication systemsoperating within the sub-6 GHz frequency band often do not meet this requirement.Additionally,the challenge of high computational complexity in distin

295、guishing static paths inthe spatial domain must not be overlooked.(2)Round-Trip Measurement MethodsThe cancellation method primarily serves to mitigate the effects of timing drift,carrier frequencyoffset,and random phase.In LOS scenarios,timing offset can be eliminated by utilizing theknown delay of

296、 the LOS path.However,in NLOS scenarios,alternative methods are required toaddress timing offset.Similar to the round-trip time(RTT)method used in 5G NR positioning,both timing offset andcarrier frequency offset between the transmitter and receiver of sensing signals can be estimatedthrough round-tr

297、ip measurements29.The underlying principle is that the movement state(position and velocity)of the sensing target remains relatively constant over a short period(e.g.,afew milliseconds to tens of milliseconds).For the same sensing target,the signal propagationdelay and Doppler frequency shift remain

298、 consistent during the round-trip measurements.Consequently,the absolute values of timing offset and carrier frequency offset are equal but haveopposite plus-minus signs,enabling their extraction or suppression.(3)Compensation MethodsCancellation methods are often limited to specific scenarios or ha

299、ve particular requirementsregarding sensing measurement configurations.For instance,delay domain cancellation requires awide signal bandwidth and the presence of a static path with high power,while spatial domain51/90cancellation requires a static path with high spatial resolution and power.These re

300、quirementsrestrict the applicability of the corresponding methods.Reference 21 systematically investigates the mechanisms through which sensing non-idealfactors affect sensing signals,revealing phenomena such as delay spectrum shift and phase driftcaused by timing drift,which subsequently lead to th

301、e delay-Doppler spectrum dispersion.Basedon these researches,the Asynchronous Delay-Doppler(ADD)method is proposed.The ADDmethod addresses delay spectrum shift and phase drift phenomena through two steps:Delay Spectrum Alignment:Oversampling IDFT is employed to obtain the delay spectrum,thereby mini

302、mizing the residual fractional shift after aligning the delay spectrum.The shiftvalues of delay spectrums between different OFDM symbols are extracted from the delayspectrums,followed by compensation for these shifts.Phase Compensation:Once the delay spectrums are aligned,phase drift resulting fromt

303、iming drift can still cause dispersion in the delay-Doppler spectrum.Phase compensationmethods include:Phase construction method:This method leverages the analytical relationship betweenphase drift and timing drift to construct the compensated phase value based on theextracted delay spectrum shift.I

304、t is straightforward to implement and is suitable forscenarios in which random phase variations are absent,such as downlink sensing oruplink sensing involving high-performance terminals.Group phase difference method:When the power of static paths dominates in the totalreceived signal power,this meth

305、od constructs the compensated phase value based on thegroup phase difference between OFDM symbols.It is applicable in scenarios whererandom phases are present.Figure 3-25 illustrates the effectiveness of the ADD method in mitigating the effects of timingdrift.The results demonstrate that the ADD met

306、hod effectively suppresses timing drift influence.Furthermore,based on the undispersed delay-Doppler spectrum acquired through ADD processing,it is straightforward to eliminate the impact of carrier frequency offset.Table 3-2 summarizes the sensing non-ideal factors that can be suppressed by the var

307、iouscancellation methods as previously discussed.The advantages and limitations of each methodhave also been outlined.52/90Table 3-2 Sensing Non-Ideal Factors Applicable to Various Sensing Non-Ideal Factors CancellationMethodsTiming offsetTiming driftCarrierfrequencyoffsetRandomphaseCancellation in

308、theantenna domain-Cancellation in thedelay domain(depending onLOS path)Cancellation in theangle domain(depending onLOS path)Round-tripmeasurement-Asynchronousdelay-Doppler-3.3.3 Bistatic Clutter Cancellation SchemesClutter refers to echo signals received by radar systems that are irrelevant to the t

309、arget.Theseclutter echoes exhibit randomness and characteristics similar to thermal noise,as they arise fromdifferent clutter objects with varying amplitudes and phases.In many cases,clutter powersignificantly surpasses the thermal noise power at the receiver.Consequently,the ability ofsensing syste

310、ms to detect targets in complicated environments relies more on the signal-to-clutterratio than the signal-to-noise ratio,making clutter cancellation essential to guarantee the sensingperformance.In radar applications,Moving Target Indication(MTI)and Moving Target Detection(MTD)arecommonly employed

311、for clutter cancellation.MTI involves suppressing stationary clutter byleveraging the Doppler shift created by target movement to emphasize moving targets30.Typically,MTI uses filters to attenuate echo signals from stationary objects,thereby enhancingthe detection performance for moving targets.In r

312、adar systems,the clutter frequency is generallyconcentrated near the zero-frequency region in frequency spectrum.For pulse-based radar,thisclutter frequency is also concentrated near zero-frequency area and the repetition frequency ofradar pulses in frequency spectrum.The fundamental principle of MT

313、I is to distinguish echosignals from targets against ground clutter by exploiting the movement characteristics of thetargets,effectively suppressing the frequency band where clutter power is concentrated.MTI oftenemploys pulse-Doppler processing to analyze echo signals from multiple continuous radar

314、 pulses,enabling the suppression of stationary clutter and the differentiation of slow-moving targets fromfast-moving ones.MTD,on the other hand,focuses on detecting mobile targets within radar53/90echoes31.It typically integrates MTI technology to enhance target detection capabilities in clutterenv

315、ironments by processing radar echo signals and utilizing a Doppler filter bank to suppressvarious forms of clutter,thus achieving both clutter cancellation and target detection.Additionally,Space-Time Adaptive Processing(STAP)technology,widely used in airborne radar,employs both spatial and temporal

316、 information from the radar receiver array to suppress clutter.Through joint adaptive filtering in both the spatial and temporal domains,STAP effectivelymitigates strong clutter and interference in airborne radar applications32.STAP technologyencompasses spatial filter design,time-domain adaptive pr

317、ocessing,and azimuth-Doppler jointprocessing.It is widely utilized in modern radar systems,particularly for addressing complicatedelectromagnetic environments and suppressing interference signals.In bistatic sensing mode,if both the transmitter and receiver are stationary,dynamic targets can beeffec

318、tively detected by canceling static clutter using classic techniques like MTI,following theelimination of non-ideal factors.However,if either the transmitter or receiver is in motion,thereflected signals from unintended targets may carry Doppler information,complicating clutteranalysis and making ca

319、ncellation more challenging.For instance,in a bistatic sensing modeinvolving a UE moving at 3 km/h while detecting a target moving at 20 km/h,the movement ofthe UE causes the echo signals of various environmental reflectors to carry Doppler information.Since the echo power from static reflectors typ

320、ically exceeds that of moving targets,the signalsfrom the moving target can easily be masked by clutter,rendering them undetectable.Asillustrated in Figure 3-27,the signal power dominated by LOS signals between transmitter andreceiver,along with echoes from other static targets in the delay-Doppler

321、spectrum,issignificantly higher than the power of the sensing targetsechoes.The Doppler region where thesensing target is located is also dominated by clutter power,making target detection difficult.Figure 3-27 Detection Results in the Delay-Doppler Domain before Clutter Cancellation54/90If there is

322、 a noticeable difference between the targets velocity and the UEs velocity,the dynamicclutter interference from UE movement can be mitigated through joint processing in the spatialand Doppler domains,enhancing the performance of moving target detection.On one hand,spatial domain filtering can be emp

323、lyed to focus on the sensing area,minimizing clutterinterference from other regions.On the other hand,since the UE moves at a relatively slow speed,the dynamic clutter power induced by its movement is concentrated in a lower Doppler frequencyregion,allowing for effective suppression through Doppler

324、filtering.To address the high echopower of static targets and minimize the impact of sidelobe signals in the Doppler domain on thesensing targets echoes,additional suppression of sidelobe power is achieved through windowing.By employing joint processing in both the spatial and Doppler domains,clutte

325、r effects can beeffectively suppressed,thereby increasing the probability of target detection.Figure 3-28 Detection Results in the Delay-Doppler Domain after Clutter Cancellation3.4 Beam Management and Precoding3.4.1 Beam Management for Bistatic SensingBeam management is essential in massive MIMO co

326、mmunications.By utilizing large-scaleantenna arrays and beamforming techniques,signal energy can be concentrated in specificdirections for both transmission and reception,thereby enhancing communication quality andcoverage.In ISAC systems,beam management is also critical for improving both communica

327、tionrates and sensing performance.Sensing beam training is closely tied to sensing services such as target detection and tracking,necessitating a well-designed beam selection and switching schemetailored to the requirementsof the sensing service.Consequently,processes and feedback information involv

328、ed in beamtraining are different for communicationand sensing.The primary distinction is that during the55/90beam management for communication,L1-RSRP,SINR,etc.are usually used as the measurementquantities of beam management.However,in sensing services,maximizing L1-RSRP and SINRof a beam pair ensur

329、es only that the energy of the received sensing signal is maximized,not thatthe sensing signal aligns with or passes through the target.For example,in Figure 3-29,beam pair2-B exhibits the highest RSRP.However,to sense the target more accurately,beam pair 1-Aprovides improved sensing performance.Thu

330、s,supporting the selection of the sensing beamnecessitates the development of a new beam training process and appropriate configurations forsensing measurements.Furthermore,the measurement results must be strongly correlated with thesensing target.Specifically,the measurement quantities should accur

331、ately indicate whether thesensing signal has been backscattered by the target,and may include metrics such as sensing SNRor the power information of signal components associated with the sensing target(the powerinformation of the path related to the sensing target).Figure 3-29 Difference Between Opt

332、imal Beam for Communication and that for SensingThe following describes the sensing beam management process,taking downlink bistatic sensing(where the base station acts as the sensing transmitter and the terminal as the sensing receiver)asan example.The terminal determines the optimal sensing beam p

333、air through the following steps:Step 1:Initial Identification of Environmental TargetsAs illustrated in Figure 3-30,during the initial access for the terminal,the base station transmits Msensing reference signals through M wide beams,while the terminal receives these signals throughN wide beams.The terminal initially identifies targets in the environment based on measurementsfrom one or more pairs