PaperlayerView exact revision on arXiv

Source-linked AI summary

An Overview of Signal Processing Techniques for Joint Communication and Radar Sensing

J. Andrew Zhang, Fan Liu, Christos Masouros, Robert W. Heath, Zhiyong Feng, Le Zheng, Athina Petropulu

arXiv:2102.12780v1eess.SP

TL;DR

JCR seeks to integrate communication and radar functions despite fundamental differences in their signal formats, system structures, and operational requirements. This paper surveys JCR signal processing from transmitter and receiver perspectives across communication-centric, radar-centric, and joint design approaches, identifying techniques and challenges for enabling integration.

  • Problem

    JCR must reconcile communication and radar requirements, including differing signal designs and the potential need for full-duplex operation.

  • Method

    The paper provides a comprehensive, balanced overview of transmitter and receiver signal processing techniques across three JCR design categories, emphasizing recent technologies.

  • Results

    The overview identifies signal processing techniques as key enablers of JCR, including sensing parameter estimation, clock-asynchrony solutions, sensing-assisted communications, and communication-rate improvement in radar-centric systems.

  • Takeaways & Limitations

    JCR can integrate communication and radar functions to improve spectrum efficiency and reduce device size, power consumption, and cost, while exploiting cooperation between the two functions.

Abstract

from arXiv · show

Joint communication and radar sensing (JCR) represents an emerging research field aiming to integrate the above two functionalities into a single system, sharing a majority of hardware and signal processing modules and, in a typical case, sharing a single transmitted signal. It is recognised as a key approach in significantly improving spectrum efficiency, reducing device size, cost and power consumption, and improving performance thanks to potential close cooperation of the two functions. Advanced signal processing techniques are critical for making the integration efficient, from transmission signal design to receiver processing. This paper provides a comprehensive overview of JCR systems from the signal processing perspective, with a focus on state-of-the-art. A balanced coverage on both transmitter and receiver is provided for three types of JCR systems, communication-centric, radar-centric, and joint design and optimization.

I. INTRODUCTION

JCR integrates communication and radar functions, motivated by shared hardware, signals, and spectrum, but their differing requirements make signal processing central. This paper surveys communication-centric, radar-centric, and jointly optimized designs, emphasizing transmitter and receiver techniques.

  • Background: Integration can reduce device size, power consumption, and cost while improving spectrum usage and potentially expanding communication and sensing capabilities.These benefits motivate applications including intelligent vehicular networks and the Internet of Things.
  • Background: JCR integrates communication and radar functions in one system, with tighter designs sharing most hardware and a single waveform optimized for both.Looser integration may use dedicated hardware or separate waveforms and offers more limited benefits.
  • JCR categories: Communication-centric designs prioritize communications and reuse communication waveforms for sensing, but sensing performance may be scenario-dependent and difficult to tune.Hardware, algorithms, and possibly communication-standard enhancements may be needed.
  • JCR categories: Radar-centric designs embed information in radar waveforms to retain near-optimal radar performance, but generally achieve limited data rates.Some radar performance loss may be tolerated to improve communications.
  • JCR categories: Joint design and optimization develops systems from the start to provide a tunable trade-off between communication and radar performance.These systems are not necessarily constrained by existing communication or radar standards.
  • Contributions: The paper surveys JCR signal processing across models, communication-centric sensing, radar-centric information embedding and reception, and joint waveform and beam optimization.It balances transmitter and receiver coverage and emphasizes receiver processing, which prior overviews had not adequately covered.

A. Beam-space Channel Models

The beam-space channel model represents multipath through delay, Doppler, angular response, amplitude, and possible timing and carrier-frequency offsets, supporting both communication and radar descriptions. Communication systems estimate composite channels, whereas radar resolves the detailed channel structure to recover sensing parameters.

  • Channel and array model: Each multipath is characterized by AoD, AoA, complex amplitude, propagation delay, Doppler frequency, and possible timing and carrier-frequency offsets.The array response uses antenna count, wavelength, spacing, and angle to model spatial propagation.
  • Channel and array model: For radar, delay, Doppler, AoA, AoD, and amplitude are sensing parameters used to determine a target’s spatial and moving information.The model relates delay to range and Doppler to velocity.
  • Communication versus radar: Communication receivers generally estimate composite channel matrices, while radar sensing must resolve detailed channel structure and estimate individual sensing parameters.This distinction motivates using a common channel model while applying different processing objectives.
  • Synchronization effects: Unlocked transmitter and receiver clocks create timing offset and CFO that can be time-varying and introduce ambiguity in radar range and speed estimation.A typical 20 PPM clock stability can produce 20 nanoseconds of timing variation over 1 millisecond, corresponding to a 6-meter ranging error.
  • Communication signal models: Single-carrier and MIMO-OFDM systems represent transmitted data or pilots through spatial precoding, while OFDM resource blocks may be discontinuous across antennas, subcarriers, and symbols.Such irregular resource allocation creates challenges for sensing parameter estimation.

C. MIMO Radar Signals and Systems

MIMO radar transmits mutually orthogonal waveforms across antennas, using pulsed or continuous-wave signals and typically low-PAPR transmission. Its signal model supports matched filtering and virtual-array spatial resolution, while MIMO-OFDM offers related orthogonal structures useful for JCR.

  • Radar waveform design: MIMO radar transmits orthogonal waveforms from different antennas, using either pulsed or continuous-wave operation.Orthogonality may span time, frequency, space, or code domains.
  • Radar waveform design: Radar transmission typically requires very low PAPR for high power efficiency, so the precoding matrix is commonly an identity matrix.The paper notes that non-identity precoding can be handled through matrix inversion when applicable.
  • Radar receiver model: Matched filtering separates the antenna waveforms because the desired waveform response is nonzero for the matching waveform and zero for the others.Stacking the matched-filter outputs produces the radar signal matrix used for subsequent processing.
  • Virtual-array processing: A monostatic MIMO radar can obtain the spatial resolution of a virtual ULA with MT MR antennas when transmitter and receiver spacings are appropriately selected.The increased aperture is meaningful when transmit and receive angles are related, although equal angles are not required.
  • MIMO-OFDM radar: MIMO-OFDM radar signals resemble communication signals but are typically non-modulated or antenna-orthogonal, making communication preambles directly usable for radar sensing.Orthogonality can be created through subcarrier allocation or time-domain codes.

2) Frequency-Hopping MIMO Radar and Frequency Agile Radar:

Frequency-hopping MIMO radar and frequency-agile radar divide bandwidth into subbands whose use changes over time, while CAESAR adds subarray-based beamforming and generalizes FH-MIMO. The section places these radar signals within broader JCR integration challenges.

  • Frequency-hopping radar: Frequency-hopping MIMO radar and frequency-agile radar use rapidly changing subsets of subbands, with hopping implemented in fast or slow time and through pulsed or continuous-wave signals.The described case is pulsed fast frequency hopping, with repeated frequency changes during a pulse repetition interval.
  • Frequency-agile radar: In FAR, all antennas share one frequency per hop and apply beamforming weights to form a steerable beam, while multi-subband extensions use several frequencies per hop.CAESAR divides the array into non-overlapped subarrays and assigns frequencies accordingly.
  • Relationship between radar schemes: CAESAR generalizes FH-MIMO by introducing beamforming capability, with FH-MIMO recovered when each subarray contains one antenna and each frequency serves one antenna.The special case is S = MT and |As| = 1.
  • Relationship between radar schemes: Both FH-MIMO radar and CAESAR are based on frequency division and can be implemented in MIMO-OFDM by applying frequency hopping to subcarriers.This connects agile radar waveform structures with communication-oriented OFDM frameworks.
  • JCR design context: C&R signals differ in application objectives, creating a JCR design challenge that requires exploiting commonalities while accounting for their differences.Radar emphasizes localization and tracking properties such as low PAPR and narrow ambiguity-function mainlobes, whereas communication signals emphasize information carrying.
  • Communication-centric JCR: Communication-centric DFRC reuses primary communication signals for sensing, including 802.11ad and mobile-network systems that use one transmitted signal for both functions.802.11ad sensing primarily exploits packet preambles, while control-PHY signals can provide wider FoV and potentially better accuracy than single-carrier or OFDM PHY signals.

B. Mobile network DFRC Systems

Mobile-network DFRC extends JCR to downlink and uplink sensing in cooperative cellular settings, where communication signals from multiple nodes can be processed for sensing. Receiver methods either use signals directly or remove data symbols through channel estimation and decorrelation.

  • Mobile-network DFRC architecture: Perceptive mobile networks define downlink and uplink sensing according to whether received downlink or uplink communication signals provide the sensing waveform.In cloud radio access networks, cooperative remote radio units can contribute downlink signals for sensing.
  • Mobile-network DFRC architecture: The multiuser MIMO-OFDMA model covers self-reflection, uplink sensing at a standalone base station, and downlink sensing at a remote radio unit.These cases differ in the contributing nodes and whether timing offsets and Doppler offsets are present.
  • Direct sensing: Direct sensing feeds received signals into sensing algorithms and can use known pilots or unknown payload data, with payload processing extending sensing capability because it lasts longer than pilots.The receiver must know or estimate the payload symbols through demodulation.
  • Direct sensing: Reorganizing subcarrier measurements into a matrix enables one-dimensional MMV compressive sensing or two-dimensional compressive sensing for delay, angle, and Doppler estimation.The two-dimensional approach can estimate delay and AoD together, while Doppler can be estimated first by stacking samples across OFDM symbols.
  • Direct sensing: Direct sensing has high computational complexity and limited applicable solutions when data symbols remain present, but it is the only option when those symbols cannot be removed because measurements are insufficient.This limitation is especially relevant when sensing uses data payloads.
  • Indirect sensing: Indirect sensing estimates channel matrices, decorrelates multiple-node signals, removes data or symbol terms, and then applies sensing parameter estimation.In multiuser MIMO, decorrelation requires at least K ≥ MT QT available symbols during a CPI and an invertible symbol matrix.
  • Indirect sensing: Indirect sensing is particularly suitable for orthogonal training and pilot symbols because matrix inversion can otherwise increase complexity and significantly enhance noise.Unknown precoding at the receiver also makes AoD estimation challenging.

2) Clutter Removal:

Clutter removal is treated as a preprocessing choice in communication environments with dense multipath: it reduces sensing complexity but may distort the signal. The section then frames sensing algorithms around harmonic-retrieval formulations and practical sampling and dimensionality constraints.

  • Clutter removal: Removing clutter before sensing can reduce the number of parameters to estimate, although it may distort the signal.The paper presents preprocessing clutter removal as generally preferable despite this distortion.
  • Sensing algorithms: Matched filtering is used in traditional radar and some DFRC systems, but its accuracy and resolution depend strongly on signal correlation properties.The paper motivates alternative methods that are less affected by communication-signal correlation properties.
  • Sensing algorithms: The decorrelated MIMO observations form a 4-D harmonic-retrieval problem that can be reduced to lower-dimensional problems and represented in matrix or tensor forms.These representations support different estimation methods and processing orders.
  • Sensing algorithms: Algorithm selection must accommodate communication signals with discontinuous, varying-interval samples and balance higher-dimensional estimation capability against computational cost.The supplied passage begins the discussion of these practical trade-offs.

D. Resolution of Clock Asynchrony

Clock asynchrony creates time-varying timing and carrier-frequency offsets that interfere with range, speed, and long-interval processing. Cross-antenna methods and mirrored MUSIC address these offsets and image ambiguities, while sensing-assisted beamforming uses predicted angles to reduce communication overhead and improve tracking.

  • Clock asynchrony: Unsynchronized transmitter and receiver clocks create time-varying timing offsets and CFO that cause range and speed ambiguity and prevent long-interval measurement aggregation.This is especially important for Doppler estimation using signals from multiple packets.
  • Clock asynchrony: Cross-antenna cross-correlation removes common timing offsets and CFO, but produces doubled terms and relative sensing parameters.The method assumes shared receiver-clock offsets across antennas.
  • Clock asynchrony: CACC-based processing requires a relatively static transmitter-receiver geometry, known transmitter location, and a dominant LOS path.The LOS contribution can be removed with high-pass filtering before estimating parameters relative to the known path.
  • Clock asynchrony: Mirrored MUSIC uses symmetry in relative delays and Doppler frequencies to halve the unknown-parameter count and simplify image-ambiguity resolution.It constructs new signals and basis vectors using original and sample-reversed versions.
  • Sensing-assisted communications: JCR predictive beamforming estimates vehicle angles from echoes, enabling continuous state estimation and high-quality communication while the vehicle remains within RSU coverage.The achievable rate relies critically on sensing and prediction accuracy for the angle of arrival.
  • Sensing-assisted communications: Using the full JCR downlink block for sensing and communication removes dedicated downlink pilots and reduces downlink overhead.The same signaling block supports beam sensing and data transmission.
  • Sensing-assisted communications: Echo-based angle estimation removes the need for uplink feedback, reducing uplink overhead and avoiding feedback quantization error.The paper links this to continuous vehicle-state estimation.

IV. JCR: RADAR-CENTRIC DESIGN

Radar-centric JCR embeds information into radar signaling while preserving radar operation as much as possible. Its principal trade-off is long-range, low-latency communication potential versus limited data rates imposed by radar waveforms.

  • Radar-centric design: Radar-centric JCR can offer communication over radar’s extraordinary range, potentially reaching hundreds of kilometers with lower latency than satellite communications.The achievable data rates are typically limited by the radar waveform.
  • Radar-centric design: Recent radar-centric DFRC work emphasizes information embedding, while communication protocols and receiver designs receive comparatively limited attention.The review focuses on MIMO-OFDM, CAESAR, and FH-MIMO radar systems.

A. Embedding Information in Radar Waveform

Radar-centric DFRC embeds data through index modulation, selecting radar-signal combinations or antenna assignments while preserving the basic waveform structure. The resulting capacity is constrained in practice by demodulation complexity and possible radar-performance effects.

  • Information embedding: Index modulation embeds information in combinations of radar parameters across space, time, frequency, or code without changing the basic radar waveform structure.For the reviewed radar families, it is implemented through frequency selection or combination and antenna selection or permutation.
  • MIMO-OFDM: MIMO-OFDM frequency combination assigns N subcarriers to MT antenna groups, with formulas depending on whether group sizes are unconstrained or equal.Equal allocation uses Ls = N/MT subcarriers per antenna.
  • CAESAR: CAESAR-based DFRC selects S of N frequencies and permutes antennas under equal virtual-subarray sizes.The number of combinations and permutations is determined by S, N, and MT.
  • FH-MIMO: For N = 8 and MT = 2, the FH-MIMO example conveys 4 bits in each information-bearing hop.The packet contains two identical preamble hops and three hops with embedded information.
  • FH-MIMO: FH-MIMO DFRC encodes information by varying the frequency set and antenna-permutation matrix across hops.The packet example uses a preamble with two identical hops followed by three information-bearing hops.
  • Practical limitations: The nominal numbers of frequency combinations and antenna permutations define maximum bit rates, while practical choices may be reduced by demodulation complexity and performance considerations.Antenna-permutation bits are described as more difficult to demodulate, and radar ambiguity-function effects also require evaluation.

B. Signal Reception and Processing for Communications

Receiver processing in FH-MIMO DFRC recovers frequency-combination and antenna-permutation patterns from sampled multantenna signals. The overview contrasts maximum-likelihood demodulation with lower-complexity sparse-recovery and DFT-based alternatives, while highlighting channel-estimation challenges and codebook effects.

  • Receiver framework: FH-MIMO DFRC receiver processing is used as a representative framework because its reception methods share features with other joint systems.The overview also incorporates MIMO-OFDM and CAESAR DFRC work.
  • Signal model: The receiver mixes, samples, and stacks signals from M_R antennas into matrix Y_k under narrowband, delay, windowing, and synchronization assumptions.Sampling at T_s = 1/B produces L_p samples per hop.
  • Demodulation: Demodulation retrieves information bits from Y_k, but maximum-likelihood detection is computationally infeasible, motivating sub-optimal compressed-sensing methods.The presentation assumes perfect synchronization and channel estimation, although channel estimation may be unnecessary for frequency-combination identification.
  • Sparse recovery: Sparse recovery identifies the M_T active frequency rows and matches their recovered coefficients to B_k to determine frequency combinations and antenna permutations.The method constructs an L_p × N dictionary and uses MMV-CS recovery.
  • DFT processing: DFT processing identifies frequency combinations by locating spectral peaks and can determine antenna permutations through channel inversion, LOS dominance, or exhaustive search.The approach is especially effective when H^-1 exists or the LOS path dominates.
  • Practical constraints: Accurate H estimation is critical but difficult because long training sequences provide certainty at the cost of affecting random frequency-hopping radar operation.A codebook’s antenna-permutation rates also depend on differences between H’s columns, while codeword constraints can reduce waveform degeneration and simplify receiver processing.

V. JCR: JOINT DESIGN AND OPTIMIZATION

Joint design and optimization gives JCR systems freedom to design communication and radar functions together rather than adapting one to an existing system. The overview therefore focuses on waveform and precoding optimization using joint communication-sensing objectives and radar-similarity constraints.

  • Scope: Jointly designed JCR systems can balance communication and radar requirements without being constrained by existing C&R systems.This category offers more signal and system design freedom than communication-centric or radar-centric approaches.
  • Waveform optimization: Joint waveform optimization is a central problem spanning multiple domains and performance metrics for communications and radar.The review mainly considers narrowband single-carrier signals, with OFDM extensions described as generally straightforward.
  • Precoding objectives: Precoding optimization can use communication or sensing objectives individually, or a weighted joint objective incorporating mutual information, waveform mismatch, or sensing-estimation accuracy.The formulation includes constraints alongside an objective function λ(P).
  • Mutual information: Mutual-information formulations combine communication capacity and radar estimation-rate objectives, including weighted sums for single-antenna OFDM and MIMO DFRC systems.A general weighted-MI formulation is concave and yields a water-filling-type solution based on channel covariance eigenvalue distributions.
  • Radar waveform properties: Communication-data randomness can undermine radar waveform properties such as correlation, PSLR, PAPR, and clutter or interference resilience.The challenge is to preserve useful radar characteristics within a joint communication-radar waveform.
  • Similarity-based design: Waveform or beampattern similarity methods approximate a benchmark radar signal while retaining communication constraints such as user SINR and transmit-power limits.The resulting covariance-matching problem is generally non-convex and may be addressed with semidefinite relaxation or manifold-based methods.
  • Direct waveform design: Direct waveform design can minimize multiuser interference relative to the intended communication symbols, producing an AWGN model when the interference vanishes.The optimization designs the DFRC waveform matrix X under communication and radar-related constraints.

3) Estimation Accuracy Based:

JCR optimization can target sensing-estimation accuracy and spatial coverage while preserving communication performance. The reviewed approaches use CRLB- or distortion-based objectives and multibeam designs for analog-array limitations.

  • Estimation accuracy: CRLB-based optimization uses lower bounds on estimation MSE because sensing signals are generally nonlinear in the parameters and direct MSE optimization lacks closed-form expressions.Pareto-optimal designs jointly improve range and velocity estimation accuracy and communication capacity in single-antenna OFDM DFRC.
  • Multibeam motivation: Analog-array beamforming makes simultaneous communication and sensing toward different directions challenging, motivating multibeam technology.The limitation is particularly relevant to mmWave JCR systems.
  • Multibeam design: Multibeam designs combine a fixed communication subbeam with a scanning sensing subbeam, allowing one signal to support both functions and extend sensing field of view.The subbeam-combination and global-optimization families can also be extended to full-digital arrays.
  • Subbeam combination: Subbeam-combination methods generate separate communication and sensing beams, shift them toward desired directions, and optimize power and phase coefficients for coherent combination.The power factor ρ and phase coefficient ϕ control the combined multibeam.
  • Constrained optimization: Multibeam constraints can maximize communication received power while enforcing sensing-direction beamforming-gain thresholds under a maximal-ratio communication combiner.The formulation specifies sensing angles and the number of gain constraints.
  • Practical use: Subbeam-combination multibeam optimization is a low-complexity, flexible approach suited to fast-changing beamforming requirements, especially in directional mmWave systems.Its practical appeal follows from simple and flexible multibeam generation and optimization.

2) Global Optimization:

Global optimization directly designs the beamforming vector under communication and sensing constraints, providing a benchmark for suboptimal multibeam methods. Related temporal-frequency optimization shows that non-uniform preambles can improve the communication-radar trade-off for sensing.

  • Global Optimization: Global multibeam optimization directly optimizes the beamforming vector while constraining waveform mismatch and sensing-direction beamforming gain.The resulting problem is generally nonconvex and NP-hard, and can be transformed into a homogeneous QCQP solved using SDR.
  • Global Optimization: 10% is the reported maximum loss in communication received signal power and beamforming waveform when subbeam combination replaces global optimization.Global optimization therefore serves as a benchmark for suboptimal multibeam schemes.
  • Signal Optimization in Other Domains: Temporal and frequency-domain signal optimization can target sensing-parameter estimation accuracy through non-uniform preambles, modified Golay sequences, and non-equidistant subcarrier placement.These designs address Doppler, range, and sidelobe-related objectives in JCR systems.
  • Signal Optimization in Other Domains: Non-uniform preambles improve velocity-estimation accuracy by optimizing preamble number and placement under a joint distortion-MMSE and CRLB objective.The objective applies log scaling to promote proportional fairness between communication and radar metrics.
  • Signal Optimization in Other Domains: Equally spaced preambles cannot effectively improve radar or communications without affecting the other, whereas non-uniform placement yields a better trade-off, particularly at large radar distances.This comparison is reported for a single-stream 802.11ad-based JCR system.
  • Conclusion: The paper concludes that JCR still faces open problems arising from differences in signal formats and system structures, with receiver processing and communication-rate improvements remaining important challenges.It identifies signal processing as central across communication-centric, radar-centric, and joint-design systems.
Loading 2102.12780v1…