Yifei Sun,Jie Li,Tong Zhang,Rui Wang,Xiaohui Peng,Xiao Han,Haisheng Tan
Abstract—In this paper,an indoor layout sensing and localization system with testbed in the 60-GHz millimeter wave(mmWave)band,named mmReality,is elaborated.The mmReality system consists of one transmitter and one mobile receiver,both with a phased array and a single radio frequency(RF)chain.To reconstruct the room layout,the pilot signal is delivered from the transmitter to the receiver via different pairs of transmission and receiving beams,so that multipath signals in all directions can be captured.Then spatial smoothing and the two-dimensional multiple signal classification(MUSIC)algorithm are applied to detect the angle-of-departures(AoDs)and angle-of-arrivals(AoAs)of propagation paths.Moreover,the technique of multi-carrier ranging is adopted to measure the path lengths.Therefore,with the measurements of the receiver in different locations of the room,the receiver and virtual transmitters can be pinpointed to reconstruct the room layout.Experiments show that the reconstructed room layout can be utilized to localize a mobile device via the AoA spectrum.
Keywords—millimeter wave,indoor sensing and localization,MUSIC algorithm,multi-input multi output(MIMO),room layout,integrated sensing and communication(ISAC),environment sensing,orthogonal frequency division multiplexing(OFDM)
Millimeter wave(mmWave)communications have been one of the key technologies of next-generation wireless networks.Despite the large bandwidth,high propagation attenuation and reflection loss of mmWave signals are the drawbacks from the communication point of view.However,these drawbacks may favor the wireless sensing performance by degrading the interference.Hence,it is of interest to exploit the sensing capability of the mmWave communication system,such that the link reliability can be improved.For example,with the location knowledge of reflectors and mobile devices,the mmWave link can be quickly recovered if link blockage occurs.
There have been a few testbeds in the existing literature designated to detect the room layout via mmWave transceiver.For example in Ref.[1],co-located mmWave transmitter and receiver were deployed at a mobile platform to detect the layout of a corridor along a planned trajectory.In Ref.[2],an mmWave indoor mapping system with co-located transceiver was proposed,which utilized orthogonal frequency division multiplexing(OFDM)radar processing to obtain sparse range-angle charts.However,the above designs may not be applicable in a wireless communication system,where the signal transmitter and receiver are separated.Moreover,the self-interference cancellation for co-located transmitter and receiver is still challenging in communication systems.It was shown in Ref.[3]that indoor ambient reflectors can be detected via the mmWave transmitter and receiver,where the transmitter is fixed and the receiver receives signals at multiple locations.In this testbed,an omnidirectional antenna on a rotation platform is used to simulate an antenna array,such that angle-of-departures(AoDs)and angle-of-arrivals(AoAs)of the propagation paths can be estimated.Moreover,based on the detected layout,the distribution of the signal to noise ratio(SNR)in the room(radio map)is also predictable.However,it is more practical to use the phased array in mmWave communication systems,where imperfect antenna elements in the phased array may degrade the estimation accuracy of AoDs and AoAs.
To quickly recover mmWave communications from link blockage,it is necessary for the access point(AP)to know both the room layout and the location of the mobile device.The localization methods in the existing literature can be classified into four categories,namely fingerprint-based localization,angle-based localization,time-of-flight-based(ToFbased)localization,and multipath-based localization,which are elaborated below.
Fingerprint-based localization:Localization based on fingerprints of received signal strength indicator(RSSI)[4,5]and channel state information(CSI)[6]has been well investigated in sub-6-GHz Wi-Fi networks.For example,RADAR[4]implemented a fingerprint-based localization system based on an offline RSSI database.In Ref.[5],a probabilistic method was proposed to improve RSSI-fingerprint-based localization.DeepFi[6]adopted CSI as the fingerprint for deep-learningbased indoor localization.In the mmWave band,instead of using fine-grained CSI or coarse-grained RSSI as the fingerprint,mid-grained beam SNR was chosen as the fingerprint in Ref.[7].However,the fingerprint-based localization method is vulnerable to a rich multipath environment.Moreover,since the fingerprints should be updated when the environment changes,collecting fingerprints at many locations could be labor-intensive.
Angle-based localization:The AoDs and AoAs of multipath can be estimated by the multiple signal classification(MUSIC)algorithm when the devices are equipped with antenna arrays.In Ref.[8],ArrayTrack estimated the AoA spectra of multiple APs from the uplink signals of the device and then generated the localization likelihood heatmap.SpotFi[9]utilized the AoA estimated by MUSIC algorithm and ToF estimated from OFDM subcarriers of direct paths to localize the target with three-antenna APs.Phaser[10]addressed the practical problems of AoA-based localization such as autocalibration and multi-AP synchronization in commodity Wi-Fi APs.However,these angle-based localization systems required multiple APs.Moreover,they applied MUSIC-based algorithms on digital multi-input multi-output(MIMO)architecture,while the analog MIMO architecture is widely used in mmWave systems.The angle estimation methods used in these sub-6-GHz systems cannot be directly applied to mmWave systems.
ToF-based localization:Chronos[11]estimated accurate ToFs of multipath by hopping between different bands.Then ToF of the line-of-sight(LoS)path was distinguished from non-line-of-sight(NLoS)paths for accurate localization.However,such frequency hopping will degrade the throughputs of other devices.ToneTrack[12]proposed a spectrum identification algorithm to discard NLoS measurement.It requested time synchronization between the transmitter and the receiver for accurate ToF estimation,and the LoS path was required for high-accuracy localization.However,the localization error might be significant when applying the ToF-based method in mmWave communication systems.This is because of the severe and time-varying carrier frequency offset between the transmitter and the receiver.
Multipath-based localization:Angle-based and ToFbased localization methods generally request multiple APs for triangular and trilateral localization.In Ref.[13],the authors combined the azimuth angles and relative ToFs of multipaths of a single AP for localization.MonoLoco[14]proposed a multipath triangulation method to achieve decimeter-level Wi-Fi localization with a single AP.However,these methods still suffer from the frequency offset as the ToF-based localization in the mmWave band.As a result,the existing localization methods for sub-6-GHz systems may not be directly applicable to mmWave systems.Furthermore,although the room layout and device location are strongly related,there is no existing work integrating the layout reconstruction and localization in mmWave communication systems.
In this paper,an indoor sensing and localization system working at 60 GHz,namely mmReality,is proposed.The mmReality consists of one transmitter and one receiver,each with one phased array.The system works in two stages,namely the layout reconstruction stage and the localization stage.In the first stage,the AoAs,AoDs,and lengths of the propagation paths between the transmitter and the receiver are estimated,where the receiver is put in multiple positions.Based on the above geometric parameters,the walls of the room can be detected to reconstruct the layout.In the second stage,the AoA spectrum for each indoor position can be predicted via the room layout,and the position of a mobile device can be detected by comparing its AoA spectrum with the prediction.
The remainder of this paper is organized as follows.In section II,the architecture of the mmReality is introduced.In section III,the estimation methods for AoAs,AoDs,and lengths of propagation paths are elaborated,and the algorithms for layout reconstruction are elaborated in section IV.Then section V describes the localization method based on room layout and AoA spectrum.Finally,the experimental results are illustrated in section VI and the conclusion is drawn in section VII.
The proposed mmReality system consists of one transmitter at a fixed location in a room and one mobile receiver,each with a single radio frequency(RF)chain and a phased array working at 60 GHz.There areNTtransmit antenna elements(TAEs)andNRreceive antenna elements(RAEs)in the phased arrays of the transmitter and receiver,respectively.The propagation paths from the transmitter to the receiver,including the LoS and NLoS paths,are illustrated in Fig.1,where the specular reflections via walls dominate the NLoS paths.By measuring the multi-carrier pilot signals with the phased array,the transmitter and receiver can cooperatively estimate the AoDs,AoAs,and lengths of propagation paths.Hence,the relative locations of the main reflectors(e.g.,walls)and the mobile receiver with respect to the transmitter can be detected via geometric relations.The receiver moves according to the planned trajectory in the room so that a complete room layout can be reconstructed.In practice,the transmitter can also communicate with multiple receivers at different locations in the room to complete the layout detection.In the following section III,we first introduce how to estimate AoDs,AoAs,and path lengths,then the layout reconstruction algorithm will be elaborated in section IV.
Fig.1 Illustration of propagation channel between the transmitter and the receiver,where the transmitter is fixed and a virtual transmitter can be observed by the receiver due to the specular reflection of the wall
A.Joint AoD and AoA Estimation
The detection of AoD and AoA is challenging with analog MIMO architecture.For example,the exhaustive beam search used in Ref.[2]is with high estimation overhead and low angular resolution.Although there has been a significant amount of research efforts spent on angle detection in multiantenna systems,e.g.,MUSIC algorithm and estimating signal parameter via rotational invariance techniques(ESPRIT)algorithm,they are designed for digital MIMO systems,where the signals at all the antenna elements can be sampled individually.Notice that in the scenario of sensing a room layout,the main reflectors(e.g.,walls)are static.Therefore,the mmWave channel can be treated as quasi-static during multiple transmissions.This has been verified and utilized in Ref.[3]for AoD and AoA detection with a single rotating antenna.In this paper,exploiting this phenomenon,the transmitter can duplicate pilot transmissions via the same transmission beams while the receiver adopts different receiving beams to resolve the individual signal on each RAE.
Specifically,the pilot signals[t](t=1,2,···,NS)is periodically transmitted viaNTdifferent analog precoders,namely{f1,f2,···,fNT},whereNSis the number of symbols in the transmission signal.For each analog precoder,the signals[t]is transmittedNRtimes,whileNRdifferent analog combiners are used at the receiver,namely{w1,w2,···,wNR}.For the elaboration convenience,we refer to the transmission with theith precoder andjth combiner as the(i,j)th transmission.The received signal of the(i,j)th transmission can be expressed as
wheredenotes the conjugate transpose of thejth combinerwj,Hdenotes the channel matrix,n[t]denotes the additive white Gaussian noise,and
denotes the TAE-to-RAE signal matrix.Aggregating the transmissions via all the precoder and combiner pairs,we have
denote the transmission and receiving array response vectors,respectively.d,λ,andxTdenote the inter-element spacing,the wavelength,and the transpose of vectorx,respectively.TheKhighest peaks off(φ,θ)refer to the estimated AoDs and AoAs of the propagation paths.
B.Path Length Estimation
OFDM ranging,also called multi-carrier ranging[11],is applied onsi[t]to estimate the length of theith path.In the OFDM system,signals are modulated to subcarriers with different frequencies for transmission.For the same distance,the phases of the received signals of all subcarriers are different,which can be exploited to estimate the path lengths.We select equally-separated subcarriers for ranging.Without the consideration of sampling frequency offset(SFO)and packet detect delay,the estimated path length can be represented by
whereϕiis the phase of theith subcarrier,denotes the range-dependent phase offset of theith subcarrier,fidenotes the frequency of theith subcarrier,cdenotes the speed of light,andLis the number of subcarriers.As a remark notice that,whendequals the ground truth,the range-dependent phase offsets of all subcarriers will be equal,i.e.,their complex exponentials are in-phase.
In practice,however,the SFO and packet detect delay are not negligible[17].As in Ref.[3],a reference calibration scheme can be used to address this issue.Suppose the measured phase isϕi,reffor theith subcarrier at a known distanced0,(14)can be reformulated as
With the technique introduced in the previous section,the geometrical parameters of propagation paths,including AoAs,AoDs,and path lengths,can be estimated.In this section,following the method introduced in Ref.[3],we first localize the mobile receiver in each measurement(referred to as the measurement points)and then reconstruct the room layout.
A.Localization of Measurement Points
As illustrated in the example of Fig.2,there are three virtual transmitters from the P1 point of view.As a remark notice that the positions of virtual transmitters depend only on the position of the transmitter and the layout of walls,and hence,the receiver should see the same set of virtual transmitters at different measurement points.
Fig.2 Localization of a measurement point,where vTx refers to the virtual transmitter
B.Layout Reconstruction
It can be observed from Fig.2 that,after the localization of virtual transmitters,the wall can be detected by drawing the perpendicular bisector between the transmitter and each virtual transmitter.To obtain the complete room layout,we can move the receiver and take measurements in a number of points,so that the virtual transmitters with respect to all the walls can be captured.
In real measurement,there are still two issues with the above reconstruction method.First,due to the measurement error,the estimated locations of one virtual transmitter detected at different measurement points may not be identical.Moreover,in addition to the first-order path(the path with one reflection from the transmitter to the receiver),some higherorder paths(the path with more than one reflection)may also be detected.As a result,the detected locations of virtual transmitters may be dispersed.To classify the estimated positions of virtual transmitters corresponding to the same wall,we use density-based spatial clustering of applications with noise(DBSCAN)method[18]and take the cluster centroid as the estimated position of the virtual transmitter.
Finally,for some complicated room layouts,there is more than one possible room layout given the locations of virtual transmitters.One example is illustrated in Fig.3,where there are two possible layouts given the same locations of 6 virtual transmitters.This ambiguity can be removed by exploiting the locations of reflection points derived in(16).For example,in Fig.3,ifAandBare the reflection points of NLoS paths from vTx2 and vTx3 to the receivers respectively,then the room profile is shown in Fig.3(a).Otherwise,ifA′andB′are the reflection points,the room profile is shown in Fig.3(b).
Fig.3 Two possible layouts with the same virtual transmitter positions due to layout ambiguity:(a)Layout with reflection points A and B;(b)Layout with reflection points A′and B′
In this section,the fast localization method for mobile receivers via AoA spectrum and reconstructed indoor layout is explained.Although the method introduced in section IV.A can be used to localize the receiver,the overhead is significant.First,the transmitter should deliver pilots dedicatedly to the receiver for at leastNTNRtimes,such that the AoA and AoD can be estimated with high resolution.However,this estimation method may not be feasible when the channel is not quasi-static,e.g.,there are moving persons in the room.Second,the offset between their local oscillators should be carefully calibrated to suppress the estimation error of path length.The cost of oscillator synchronization could be high,especially in the mmWave band.
In practice,the transmitter(i.e.,the AP)would periodically broadcast control information to all directions,so that each receiver can detect all the potential AoAs of the signals from the transmitter via periodic beam search.Hence,raw AoA estimation can be made by the receiver.By matching the observed dominant AoAs with the locations in the reconstructed room layout,the mobile receiver can be pinpointed.This facilitates localization without dedicated signaling overhead.The existing works exploiting the AoA spectrum(including the AoAs and signal powers at the arrival directions)in localization usually rely on multiple APs.Compared with the existing works in sub-6-GHz,we shall show single transmitter might be sufficient to generate the AoA spectrum for localization in the mmWave band,with the assistance of virtual transmitters.The reasons are elaborated on below.First,it is shown by experiments that the propagation paths arrived at the receiver are dominated by the LoS and first-order NLoS paths,leading to the limited number of virtual transmitters.Second,the phased array at the mmWave band is of smaller size,so that it can be implemented on mobile devices,and used to resolve the real and virtual transmitters in different directions.
Specifically,we choose|G|positions as the possible position in the reconstructed room layout,denoted asG={1,2,···,|G|}.In practice,the set of the feasible positionsGcan be generated by mesh grids.Then the expected AoA spectra at all positions inGis calculated according to the positions of the transmitter and virtual transmitters.LetΦi={φi,j|j=1,2,···}be the set of AoAs for the positioni,andbe the set of AoAs observed at the real receiver.We define the error function of an expected AoA profile with respect to the measured one as
whereAthis a constant threshold,and|Φi|denotes the cardinality of the setΦi.Then the estimated position with the measured AoA profilecan be obtained by finding the position with the smallest value of the error function out of all feasible positions in the reconstructed layout.
The blessing11 to Karen and I was twofold. In addition to ending years of confusion and estrangement12（，） , we learned to share the joy of being the proud daughters of two extraordinary fathers.
As a remark notice that the thresholdAthin(18)is to avoid a significant difference in the error function when the desired LoS and first-order reflection paths are blocked,or higherorder reflection paths are captured in the AoA detection.
A.mmReality System Implementation
The block diagram of mmReality is illustrated in Fig.4.Taking the transmitter side as an example,one softwaredefined radio(SDR)generates the baseband signal and upconverts it to the intermediate frequency(IF)band centered at 500 MHz.The IF signal is fed into 90◦and 180◦power splitters sequentially to generate differential in-phase and quadrature(IQ)signals.Then the IF signal is further up-converted to 60 GHz and finally transmitted by a 16-antenna phased array.The antenna selection,beam switch,and power gain of the phased array are controlled by a host computer.To achieve sufficient ranging accuracy,the transmission signal is modulated with 8 subcarriers and 12.5 MHz bandwidth via OFDM technology.
Fig.4 Hardware architecture
Before the measurements,the inter-element spacing and the inter-element phase offset of both phased arrays at the transmitter and the receiver should be calibrated,which is elaborated below.
Inter-element spacing calibration:Although antenna arrays are designed with half-wavelength inter-element spacing to generate beam patterns with a single main lobe and low side lobes,the wavelength varies at different carrier frequencies.The difference of adjacent carrier frequencies in the 60 GHz band is generally more than 1 GHz and cannot be neglected.Moreover,the direct measurement of inter-element spacing may not be accurate due to the short wavelength and small antenna size[19].
To address the above issue,we measure the inter-element spacing via the received signals of the phased array.We use a transmitter with a horn antenna to transmit a single tone and mount the phased array to be calibrated on a rotation platform as the receiver.The mmWave absorbers are used to suppress the potential NLoS paths.The received signals are collected atNθdifferent AoAs.For a certain AoA,each antenna element is triggered alternatively to receive the same single tone.Letϕm,nbe the estimated phase of thenth antenna element at themth AoA,denoted asθm,then the inter-element spacing can be estimated by
Inter-element phase offset calibration:Notice that the phase offset is caused by the different transmission lines of antenna elements,we calibrate it with a transmitter at the boresight direction of the phased array.In the above calibration of inter-element spacing,letθ1=0,the phase offset difference of two adjacent elements(say between thenth and(n−1)th antenna elements)due to transmission lines can be estimated directly byϕ1,n−ϕ1,n−1.As a result,the phase offset of all the remaining antenna elements with respect to the first element can be compensated.
The mmReality system is deployed in a corridor with an irregular layout as illustrated in Fig.5.The position of the transmitter is fixed,and the receiver is put at 15 measurement points respectively as in Fig.8.At each measurement point,the OFDM signal is transmitted via 16×16 pairs of transmission and receiving beams.
Fig.5 Experimental environment of the corridor
Fig.8 Corridor layout illustration with the positions of the transmitter and measurement points
B.Accuracy of AoA/AoD Estimation
In this part,the accuracy of AoA and AoD estimations based on the MUSIC algorithm is illustrated.The estimated AoAs/AoDs and the ground truths for LoS paths and NLoS paths(first-order reflection paths)are illustrated in Fig.6(a)and Fig.6(c),respectively.The ground truth AoAs/AoDs in the environment is measured by a laser rangefinder.Since the AoA/AoD in 360◦azimuth can be estimated by rotating the phased array towards the four quadrants,we only estimate the angles ranging from−45◦to 45◦.It can be observed that the estimated AoAs/AoDs match the ground truth in both LoS and NLoS scenarios.Moreover,the cumulative distribution functions(CDF)of the magnitudes of the estimation errors for the LoS scenario and the NLoS scenario are illustrated in Fig.6(b)and Fig.6(d),respectively.It can be observed that 90%of the estimations are with estimation errors less than 4◦.
Fig.6 Illustration of AoA/AoD estimation:(a)Estimated AoA/AoD versus ground truth in LoS scenario;(b)CDF of AoA/AoD estimation error in LoS scenario;(c)Estimated AoA/AoD versus ground truth in NLoS scenario;(d)CDF of AoA/AoD estimation error in NLoS scenario
C.Accuracy of Path Length Estimation
Fig.7(a)and Fig.7(c)illustrate the estimated path lengths and the ground truth for LoS paths and NLoS paths(first-order reflection paths),respectively.The estimated path lengths range from 1.6 m to 10 m for a typical indoor scenario.It can be observed that the estimated path lengths match the ground truths in both LoS and NLoS scenarios.The CDFs of the path length estimation errors for LoS paths and NLoS paths are illustrated in Fig.7(b)and Fig.7(d),respectively,where the average estimation error for LoS paths and NLoS paths are 0.15 m and 0.25 m,respectively.It is intuitive that the estimation for LoS paths performs better than that for NLoS paths because NLoS paths suffer severe reflection loss.
Fig.7 Illustration of path length estimation:(a)Estimated path length versus ground truth in LoS scenario;(b)CDF of the estimation error of path lengths in LoS scenario;(c)Estimated path length versus ground truth in NLoS scenario;(d)CDF of the estimation error of path lengths in NLoS scenario
D.Room Layout Reconstruction
Integrating the estimated AoAs,AoDs,and path lengths at all measurement points,the reconstructed layout of the corridor is illustrated in Fig.9,where the blue solid line and the red dotted line represent the real and estimated walls,respectively.It can be seen that all six walls are detected with high estimation accuracy,as both lines almost overlap.Moreover,the estimated positions of measurement points,virtual transmitters,and reflecting points are also illustrated in Fig.9.As a remark notice that higher-order reflections can be found in the measurements,although the LoS and first-order reflection paths are dominant.However,since the virtual transmitters of the higher-order reflection paths are sparse,they are eliminated by the DBSCAN algorithm.Hence,only the estimated virtual transmitter positions of the first-order reflection is shown in Fig.9.
Fig.9 Corridor layout reconstruction
The CDFs of the localization errors of the measurement points and reflection points are illustrated in Fig.10(a)and Fig.10(b),respectively.The average localization error of the measurement points is 0.42 m,and 90% of the localization errors are below 0.8 m,while those of the estimated reflection points are 0.6 m and 1.2 m respectively.The localization error of reflection points is generally larger.This is because the localization of reflection points is based on the estimated positions of measurement points.
Fig.10 CDF of the localization error:(a)Measurement points;(b)Reflection points
E.Localization Accuracy via AoA Spectrum
Based on the detected corridor layout,the localization performance via the AoA spectrum is demonstrated in this part.Specifically,the receiver is randomly deployed at 25 positions in the corridor successively,and the AoA spectrum is measured by the receiver at each position for localization.The CDF of localization error is illustrated in Fig.11.It can be observed that 90% of the localization error is within 1.3 m,and the average localization error is 1.0 m.Two examples of localization via AoA spectrum are shown in Fig.12,where different colors are used to demonstrate the value of the error functionfdefined in(18).The localization error mainly results from the localization error of virtual transmitters.
Fig.11 CDF of the localization error based on the reconstructed layout and measurements of AoA spectrum
Fig.12 Two examples of localization via AoA spectrum
F.Impact of Measurement Point Number
We evaluate the impact of the number of measurement points on the localization accuracy in another indoor environment as illustrated in Fig.13(a).The detection accuracy of the marked wall(as well as the virtual transmitter via the wall)is evaluated.As shown in Fig.13(b),the transmitter is deployed 4.06 meters away from the wall and 21 points between the transmitter and the wall are taken as the candidate measurement points.The localization accuracy of the virtual transmitter versus the number of measurement points is shown in Fig.14.For a given number,the measurement points are randomly picked from the 21 candidates 100 times for averaging.It can be observed that with more than 16 measurement points,the localization error can converge to 0.26 m.
Fig.13(a)Experimental environment of the office;(b)Office layout and the positions of the transmitter and measurement points
Fig.14 Localization error versus the number of measurement positions
G.Impact of Measurement Time
We then analyze the sensitivity of the virtual transmitter localization with respect to the measurement time,i.e.,the number of received OFDM symbols for each pair of precoder and combiner.As shown in Fig.15,the number of received OFDM symbols will affect the accuracies of both angle and path length estimation,and finally the estimated position of the virtual transmitter.The localization error decreases significantly with the OFDM symbol number at the very beginning.With more than 40 received OFDM symbols,the localization error of the virtual transmitter will be below 0.4 m.Note that it takes around 13.1 ms to transmit 40 OFDM symbols for all the pairs of precoder and combiner.
Fig.15 Localization error versus the number of received OFDM symbols
In this paper,the mmReality system for indoor layout reconstruction and fast localization was elaborated.Exploiting the quasi-statistic channel,the 2D-MUSIC algorithm was used to detect the AoAs and AoDs of the paths between the transmitter and the receiver with an analog MIMO front-end.The path length can then be estimated via multi-carrier ranging.Based on the AoAs,AoDs,and path lengths estimated by the receiver at different locations,the indoor layout can be reconstructed.With the layout knowledge,we continue to show that the receiver can be localized via the observed AoAs.The experiment results of this paper demonstrated the feasibility to track the environment and trajectory of mobile devices via mmWave communication signals.With both environment and trajectory information,communication efficiency may be improved,which is a promising topic for future study.
In our current testbed,the mmWave RF front-end support IEEE 802.11ad communication,i.e.,gigahertz bandwidth.However,due to the limited sampling rate of the baseband processor and throughput of the network interface card,we can only apply 12.5 MHz bandwidth at 60 GHz for wireless sensing.The limited bandwidth in the baseband results in low-range resolution(12 meters for 12.5 MHz bandwidth)for room layout reconstruction.The system design supporting gigahertz-bandwidth mmWave communications and sensing is left for future work.
推荐访问:Sensing Localization Indoor