PaperlayerView exact revision on arXiv

Source-linked AI summary

IEEE 802.11ad-based Radar: An Approach to Joint Vehicular Communication-Radar System

Preeti Kumari, Junil Choi, Nuria Gonzalez-Prelcic, Robert W. Heath

arXiv:1702.05833v1cs.IT

TL;DR

Existing approaches achieve at most 27 Mbps, below the requirements of some applications. This paper develops an IEEE 802.11ad-based mmWave automotive radar using the standard's preamble and radar algorithms, achieving Gbps data rates alongside accurate range and velocity estimation.

  • Problem

    Existing approaches achieve data rates of at most 27 Mbps, much less than the requirements for some applications.

  • Method

    The paper develops an IEEE 802.11ad-based mmWave automotive radar that exploits repeated Golay complementary sequences in the standard's preamble and uses single- and multi-frame radar algorithms.

  • Results

    Gbps data rates are achieved for a CPI of 0.06 ms or more, while velocity estimation reaches the desired performance even at SCNR as low as -20.5 dB for a CPI of 4.2 ms.

  • Takeaways & Limitations

    The results indicate that IEEE 802.11ad can enable a joint mmWave vehicular communication-radar system with higher-than-minimum range resolution and accuracy.

Abstract

from arXiv · show

Millimeter-wave (mmWave) radar is widely used in vehicles for applications such as adaptive cruise control and collision avoidance. In this paper, we propose an IEEE 802.11ad-based radar for long-range radar (LRR) applications at the 60 GHz unlicensed band. We exploit the preamble of a single-carrier (SC) physical layer (PHY) frame, which consists of Golay complementary sequences with good correlation properties, as a radar waveform. This system enables a joint waveform for automotive radar and a potential mmWave vehicular communication system based on IEEE 802.11ad, allowing hardware reuse. To formulate an integrated framework of vehicle-to-vehicle (V2V) communication and LRR based on a mmWave consumer wireless local area network (WLAN) standard, we make typical assumptions for LRR applications and incorporate the full duplex radar assumption due to the possibility of sufficient isolation and self-interference cancellation. We develop single- and multi-frame radar receiver algorithms for target detection as well as range and velocity estimation within a coherent processing interval. Our proposed radar processing algorithms leverage channel estimation and time-frequency synchronization techniques used in a conventional IEEE 802.11ad receiver with minimal modifications. Analysis and simulations show that in a single target scenario, a Gbps data rate is achieved simultaneously with cm-level range accuracy and cm/s-level velocity accuracy. The target vehicle is detected with a high probability of detection ($>$99.9$\%$) at a low false alarm of 10$^{-6}$ for an equivalent isotropically radiated power (EIRP) of 43 dBm up to a vehicle separation distance of 200 m.

I. INTRODUCTION

The paper proposes an IEEE 802.11ad-based mmWave joint vehicular communication-radar system that reuses a standard SC PHY waveform and shared hardware. Simulations indicate simultaneous Gbps communication and long-range radar performance with cm-level range accuracy, cm/s-level velocity accuracy, and reliable detection.

  • Motivation: Vehicular applications require Gbps data rates, while existing DSRC achieves at most 27 Mbps.The paper motivates mmWave spectrum as a route to higher-rate connected vehicles.
  • Related work: Most prior joint communication-radar systems use waveforms that are not based on a communication standard.Earlier systems are classified as simultaneous or non-simultaneous, with OFDM approaches facing sidelobe and PAPR limitations.
  • Proposed system: The proposed system uses the IEEE 802.11ad SC PHY frame for both automotive radar and V2V communication, enabling shared spectrum and hardware reuse.Its mmWave standard waveform provides large bandwidth for communication and radar operation.
  • Methods: Single- and multi-frame algorithms use the IEEE 802.11ad preamble with conventional WLAN synchronization and channel-estimation techniques for detection, range, and velocity estimation.The framework includes single- and multi-target processing under typical LRR and full-duplex assumptions.
  • Results: For multiple targets, the system achieves < 0.1 m range resolution and < 0.6 m/s velocity resolution using multiple frames in a 4.2 ms CPI.The paper also reports estimation MSEs close to their CRLBs, with a range-MSE difference less than 2 cm2 attributed to WLAN symbol synchronization accuracy.

II. THE IEEE 802.11AD PREAMBLE

The paper evaluates the IEEE 802.11ad SC PHY preamble as a radar waveform, using its STF and CEF for synchronization, channel estimation, and target parameter estimation.

  • Preamble structure: The SC PHY preamble contains an STF and CEF whose communication synchronization and channel-estimation functions support radar processing.The CEF contains a 512-sample Golay complementary pair, while the STF uses repeated 128-sample Golay sequences.
  • Preamble structure: The preamble findings based on SC PHY modulation can be extended to other IEEE 802.11ad PHY frames.The SC PHY preamble is similar to preambles in other IEEE 802.11ad PHY frames.
  • Preamble structure: The STF supports frame synchronization and frequency-offset estimation, which are leveraged for target range and velocity estimation.The paper uses frame synchronization for range estimation and frequency-offset estimation for velocity estimation.
  • Preamble structure: The CEF is used to estimate communication-channel parameters, and the same channel-estimation algorithm can support target range and velocity estimation.Radar algorithms use the STF and CEF jointly or use the CEF after the STF.
  • Preamble structure: STF-and-CEF joint processing supports a longer operating range than processing that uses the CEF after the STF.CEF-after-STF processing retains the perfect autocorrelation property of Golay complementary sequences.

B. Ambiguity Function

The IEEE 802.11ad Golay preamble is assessed through its ambiguity function and embedded in a joint V2V communication-radar scenario for estimating target range and velocity.

  • B. Ambiguity Function: The 512-sample Golay complementary pair has perfect zero-Doppler autocorrelation with no sidelobes, supporting radar target detection.The paper contrasts this property with FMCW signals typically used in long-range radar.
  • B. Ambiguity Function: The Golay pair is less tolerant of large Doppler shifts, but remains acceptable for long-range radar under small normalized vehicular Doppler shifts.The ambiguity-function diagram uses delay τ, symbol period Ts, Doppler shift ν, and Tp = 512Ts.
  • B. Ambiguity Function: A source vehicle transmits one IEEE 802.11ad waveform for both V2V communication and radar, using echoes to estimate target range and velocity.The radar receiver is mounted on the source vehicle, while the communication receiver is on the recipient vehicle.
  • B. Ambiguity Function: The system assumes narrow, aligned beams, no blockage, a single data stream, and quasi-stationary target conditions during a coherent processing interval.The TX and RX beamforming vectors are incorporated into the baseband model and remain time invariant within the interval.

B. Channel and Target Models

The channel model combines LOS-dominated mmWave V2V communication with beamformed radar operation, while explicitly modeling assumptions about beams, interference, mobility, and propagation.

  • B. Channel and Target Models: Uniform planar arrays and analog beamforming provide directional processing for the joint communication-radar system.The model incorporates TX and RX beamforming vectors and uses steering vectors indexed by azimuth and elevation.
  • B. Channel and Target Models: Full-duplex operation is assumed possible through antenna separation, beamforming, circulators, and self-interference cancellation.The IEEE 802.11ad medium-access protocol is also assumed to avoid inter-user interference from other vehicles.
  • B. Channel and Target Models: The source and recipient vehicles are represented with aligned TX/RX beams, no blockage, and a recipient vehicle modeled as a single point target.The target is assumed to have constant range, direction, and relative radial velocity during a coherent processing interval.
  • B. Channel and Target Models: The communication channel is modeled as a one-way, frequency-flat, LOS-dominated Rician channel between the source and recipient vehicles.The model includes a large-scale path-loss gain and a small-scale channel component.
  • B. Channel and Target Models: The path-loss exponent depends on the vehicular scenario, although values close to 2 are associated with LOS outdoor urban and rural mmWave channels.Numerical simulations study the effect of the path-loss exponent on radar and communication performance.

2) Target and Clutter Model:

The radar model represents target echoes and clutter in a delay-Doppler framework, then applies low-complexity pulse-Doppler processing to estimate target delays and Doppler shifts.

  • 2) Target and Clutter Model:: The model includes dominant direct-path target echoes together with multipath spread-Doppler clutter.The radar channel is treated as a doubly selective time- and frequency-selective model.
  • 2) Target and Clutter Model:: Each target path is characterized by direction, round-trip delay, small- and large-scale gains, and Doppler shift.Range and relative velocity map to delay and Doppler through τp = 2ρp/c and νp = 2vp/λ.
  • 2) Target and Clutter Model:: A single-target model uses Np = 1, whereas Np > 1 represents multiple targets, with the direct source-recipient path treated as the zeroth path.Far targets are assumed to satisfy ρp ≫ vp/T, allowing constant path gain during the coherent processing interval.
  • 2) Target and Clutter Model:: Classical low-complexity pulse-Doppler algorithms operate on the delay-Doppler map to estimate target delays and Doppler shifts.The effective multi-target model is nonlinear in the physical parameters, motivating the linear delay-Doppler representation.
  • 2) Target and Clutter Model:: The delay-Doppler map discretizes delay and Doppler at resolutions Δτ = 1/W and Δν = 1/T.It uses uniformly spaced delays τℓ = ℓ/W and Doppler shifts νd = d/T.

C. Received Signal Model

The model describes communication and radar signals over a coherent processing interval containing multiple frames, using matched filtering, synchronization, sampling, and target echoes with noise and clutter.

  • The coherent processing interval contains M frames, with each frame modeled as K samples.
  • The received communication signal is represented after matched filtering, time/frequency synchronization, and symbol-rate sampling, with additive white Gaussian noise.
  • The radar signal model uses a stop-and-hop assumption, so the echo delay corresponds to target range at the beginning of each pulse.
  • The radar model includes target-dependent phase modulation, complex AWGN, and clutter modeled as an IID complex Gaussian process.
  • The single-target received-signal model extends to multiple targets by adding corresponding target terms.
  • The processing framework leverages STF and CEF information from multiple frames within a CPI.

IV. PROPOSED RECEIVER PROCESSING TECHNIQUES FOR ENABLING RADAR FUNCTIONS

The proposed receiver reuses IEEE 802.11ad communication processing for radar detection, range estimation, and velocity estimation, extending single-frame operations across multiple frames.

  • The receiver combines communication and radar modules at the source vehicle for joint IEEE 802.11ad communication-radar processing.
  • The radar module performs CFAR vehicle detection, time-synchronization-based range estimation, and frequency-synchronization-based velocity estimation.
  • The algorithms extend communication processing from a single frame to multiple frames within a CPI, requiring minimal receiver modifications.
  • Timing processing uses STF-based coarse synchronization, CFO estimation, and CEF-based fine synchronization and channel estimation.
  • Single-frame communication-module Doppler estimation is inaccurate at low SNR because of small Doppler shifts and limited integration time, motivating multi-frame radar estimation.
  • The energy-based symbol synchronization and STF/CEF peak-detection methods estimate fractional delay and perform well even at low SCNR.

B. Single Target Radar Processing per CPI

The radar module detects targets with CFAR and estimates range from preamble timing and correlation, using STF or CEF processing to obtain coarse and fine delay estimates.

  • CFAR detection thresholds are applied to either the channel estimate or received–transmitted preamble cross-correlation energy.
  • Using the entire preamble increases probability of detection at a given false-alarm probability but raises sidelobes, which is unfavorable for multi-target detection.
  • The preamble is used for single-target detection, while the CEF is used for multi-target detection when SNR is high.
  • Coarse range estimation uses frame-start detection, while fine range estimation uses symbol-boundary detection or STF/CEF peak detection with fractional-delay correction.
  • 0.1 m range accuracy is met because fine delay estimation has an error below one sample, satisfying the LRR specification.
  • 0.8 mm range accuracy is obtained in simulations with the RRC pulse-shaping filter using a single IEEE 802.11ad frame.

3) Velocity Estimation:

Velocity is estimated from Doppler shifts using single- or multi-frame frequency-offset processing, with multiple frames improving accuracy through longer integration but creating accuracy–ambiguity trade-offs.

  • The receiver estimates target relative velocity by estimating the Doppler shift of the target echo.
  • Single-frame Doppler estimation may fail to meet the desired LRR velocity accuracy because its integration time is small.
  • A multi-frame Moose-based algorithm is proposed to improve Doppler frequency estimation across the CPI.
  • Multiple frames improve frequency-offset estimation relative to a single frame because they provide larger integration time.
  • Velocity-estimation accuracy improves with more preambles and higher radar SCNR, while increasing training duration reduces communication data rate.
  • The multi-frame approach creates a trade-off between velocity accuracy and communication data rate, and between accuracy and the unambiguous velocity span.

C. Multi-target Radar Processing per CPI

The multi-target radar processing forms a delay-Doppler map from channel estimates across frames, then detects and estimates each target from delay and Doppler bins. Simulations evaluate detection, estimation accuracy, and the communication–velocity trade-off under automotive radar conditions.

  • Processing: The receiver computes a delay-Doppler map by applying an M-point DFT to zero-padded channel estimates, then thresholds it to detect multiple targets.Each detected target’s range and velocity come from its corresponding delay and Doppler bins.
  • Processing: More frames within a CPI improve velocity resolution because the velocity-resolution bound decreases as the integration duration increases.The paper states that velocity-estimation resolution increases with the number of frames.
  • Resolution: Less than 8.52 cm range resolution and less than 0.6 m/s velocity resolution are theoretically achievable with W > 1.76 GHz and T > 4.2 ms.The proposed algorithms achieve these bounds in numerical results.
  • Detection: PD exceeds 99.9% at PFA = 10^-6 when received SCNR exceeds -20.5 dB.Monte Carlo evaluation uses 10,000 trials and compares several false-alarm probabilities.
  • Limitations: At SCNR below 10 dB, double-frame processing still fails to achieve the desired 0.1 m/s velocity accuracy.The paper motivates using multiple frames within a CPI, while noting that performance depends on the number of frames.
  • Communication–radar trade-off: Using more frames improves velocity accuracy but reduces communication data rate for a fixed CPI duration.Gbps communication and cm/s-level velocity estimation are simultaneously achieved for a CPI of 0.06 ms or more.
  • Range estimation: Single-frame amplitude-based STF/CEF peak detection achieves better than 0.1 m range accuracy, with MSEs within 2 cm^2 of the CRLB.The phase-based CEF symbol-boundary method exceeds the desired accuracy only above 6 dB SCNR, while STF/CEF peak detection does so above 0 dB.

B. Multi-target Scenario

The multi-target scenario estimates two vehicles from a normalized delay-Doppler map generated over a multi-frame CPI. The results show strong range separation but substantially poorer Doppler resolution for the short CPI, which improves with longer integration.

  • Resolution: The short-CPI Doppler response has a wide mainlobe and high sidelobes, yielding a velocity resolution of 34.375 m/s.The broad Doppler mainlobe limits velocity resolution in the simulated multi-target case.
  • Scenario and map: For a 0.072 ms CPI using 10 frames, the delay resolution is 0.07 m and the Doppler resolution is 13,750 KHz.The map is normalized to the maximum received power and interpolated along the Doppler dimension for visualization.
  • Scenario and map: The simulated delay-Doppler maps contain two dominant vehicle reflections separated in delay, corresponding to a 4.26 m range difference.Vehicle responses appear in distinct delay bins, while Doppler responses occupy the first and second Doppler bins.
  • Resolution: Increasing CPI duration beyond 4.2 ms improves velocity resolution to less than 0.6 m/s, while the range resolution remains 7 cm.The paper attributes the velocity-resolution dependence to CPI duration and reports low sidelobes along the range dimension.

VI. CONCLUSIONS

The paper develops an IEEE 802.11ad-based mmWave automotive radar that reuses standard WLAN receiver processing and supports joint communication-radar operation. Single- and multi-frame algorithms provide range and velocity estimation, with simulations showing high accuracy alongside Gbps communication rates.

  • The proposed radar uses the IEEE 802.11ad SC PHY preamble, consisting of repeated Golay complementary sequences, as its radar waveform.The waveform is exploited because of its perfect auto-correlation property at zero Doppler shift.
  • Single- and multi-frame algorithms support single- and multi-target detection together with range and velocity estimation.The algorithms leverage standard WLAN receiver and classical pulse-Doppler radar processing techniques, evaluated analytically and through simulations.
  • The target vehicle is estimated with higher range resolution and accuracy than the LRR requirements of 0.5 m range resolution and 0.1 m range accuracy.
  • Less than 0.1 m/s velocity accuracy is achieved at SCNR as low as -20.5 dB for a CPI of 4.2 ms using multi-frame processing.The same CPI provides velocity resolution below 0.6 m/s.
  • Gbps communication data rates and the desired velocity accuracy are simultaneously achieved for a CPI of 0.06 ms or more.The paper also reports a trade-off between velocity estimation accuracy and communication data rate for fixed CPI duration.
  • The results indicate that an IEEE 802.11ad-based joint communication-radar system is a promising option for next-generation automotive applications.Future work considers waveform modifications that adapt the trade-off between radar estimation accuracy or resolution and communication data rate to vehicular scenarios.

APPENDIX A

The appendix derives a Cramér–Rao lower bound for velocity estimation from multiple IEEE 802.11ad SC PHY preambles. It formulates the received training samples and Fisher information matrix used to characterize Doppler estimation accuracy.

  • The velocity-estimation CRLB is derived using the preamble of multiple IEEE 802.11ad SC PHY frames.
  • The received training sequence is indexed across training-symbol locations and frames, with frame spacing represented by K samples.The received samples include signal energy, channel response, Doppler phase, and noise under perfect symbol synchronization.
  • The Fisher information matrix is computed for the parameter vector containing received energy, Doppler angular frequency, and channel phase.The CRLB for Doppler frequency is related to ω0 = 2πν0Ts.
  • For consecutive samples in a single frame, the appendix gives a CRLB expression under the stated high-radar-SNR condition.
  • For non-consecutive training samples and a large number of frames, the appendix simplifies the CRLB expression for multi-frame processing.
Loading 1702.05833v1…