Performance Analysis of Millimeter Wave Radar Waveforms for Integrated Sensing and Communication

Next-generation intelligent transportation systems require both sensing and communication between road users. However, deploying separate radars and communication devices involves the allocation of individual frequency bands and hardware platforms. Integrated sensing and communication (ISAC) offers a robust solution to the challenges of spectral congestion by utilizing a shared waveform, hardware, and spectrum for both localization of mobile users and communication. Various waveforms, including phase-modulated continuous waves (PMCW) and frequency-modulated continuous waves (FMCW), have been explored for target localization using traditional radar. On the other hand, new protocols such as the IEEE 802.11ad have been proposed to support wideband communication between vehicles. This paper compares both traditional radar and communication candidate waveforms for ISAC to detect single-point and extended targets. We show that the response of FMCW to mobile targets is poorer than that of PMCW. However, the IEEE 802.11ad radar outperforms PMCW radar and FMCW radar. Additionally, the radar signal processing algorithms are implemented on Zynq system-on-chip through hardware-software co-design and fixed-point analysis to evaluate their computational complexity in real-world implementations.

💡 Research Summary

The paper investigates integrated sensing and communication (ISAC) for next‑generation intelligent transportation systems (ITS) by comparing four millimeter‑wave (mmWave) waveforms that can simultaneously support radar sensing and high‑rate data transmission. The authors focus on the 60 GHz band, where both automotive radar and vehicle‑to‑everything (V2X) communication can coexist, alleviating the spectral congestion that arises when separate sub‑6 GHz communication links and 77 GHz radars are deployed.

The four candidate waveforms are: (1) conventional frequency‑modulated continuous wave (FMCW), (2) phase‑modulated continuous wave (PMCW) where communication symbols are embedded in the phase of the carrier, (3) a standard Golay complementary sequence taken from the IEEE 802.11ad preamble, and (4) a Doppler‑resilient Golay sequence derived from the Prouhet‑Thue‑Morse construction. All waveforms share identical system parameters—sampling rate, pulse repetition interval (2 µs), coherent processing interval (4 ms), bandwidth (1.76 GHz), and carrier frequency (60 GHz)—ensuring a fair comparison of range resolution (0.085 m), maximum unambiguous range (44 m), velocity resolution (0.3 m/s), and maximum unambiguous velocity (625 m/s).

Two realistic target models are simulated: a pedestrian composed of 27 point scatterers (average RCS ≈ 0 dBsm) moving at 2 m/s, and a midsize car modeled with 6 905 metallic plates (average RCS ≈ 10 dBsm) moving at 10 m/s. The targets are line‑of‑sight and multipath effects are ignored to isolate waveform performance.

Range‑Doppler (RD) maps generated from the simulated echoes reveal distinct behaviors. FMCW exhibits the poorest autocorrelation for moving targets, producing high sidelobes and a peak‑to‑sidelobe ratio (PSLR) of only ~8 dB. PMCW improves the PSLR to ~13 dB but still suffers noticeable sidelobes that blur the spatial extent of extended targets. The standard IEEE 802.11ad Golay sequence provides near‑perfect autocorrelation for static objects, yet its PSLR degrades (≈20 dB) when the target moves because Doppler‑induced phase shifts break the complementary property. The Doppler‑resilient Golay sequence restores the complementary behavior even under motion, achieving a PSLR of ~43 dB and virtually zero range sidelobes for both pedestrian and car scenarios. Consequently, the Doppler‑resilient Golay waveform delivers the most accurate range and velocity estimates for moving, extended targets.

Beyond algorithmic performance, the authors implement the radar signal processing (RSP) chain on a Xilinx Zynq MPSoC (ZCU111) platform using hardware‑software co‑design (HSCD). Initially, the entire processing (including matched filtering) runs on the ARM Cortex‑A53 processing system (PS) with double‑precision floating‑point (DPFL). Offloading the matched‑filter block to programmable logic (PL) and switching to single‑precision floating‑point (SPFL) yields a speed‑up of up to 10.6× compared with the PS‑only implementation, at the cost of increased DSP, LUT, and BRAM usage. To further reduce resource consumption, a fixed‑point (FP) design with 24‑bit word length is explored. The FP implementation cuts DSP and LUT utilization by up to 30 % and power consumption by up to 9 % while preserving the PSLR and detection performance of the SPFL version.

The hardware results confirm that the Doppler‑resilient Golay waveform not only outperforms FMCW and PMCW in detection quality but also maps efficiently onto edge‑computing hardware. The fixed‑point FPGA implementation demonstrates that real‑time ISAC processing is feasible on automotive‑grade SoCs without sacrificing accuracy, offering a practical pathway for integrating mmWave communication and radar sensing in future vehicles.

In summary, the study establishes that (i) for moving targets, the Doppler‑resilient Golay sequence embedded in IEEE 802.11ad provides superior range‑Doppler performance with negligible sidelobes; (ii) traditional FMCW and PMCW waveforms are inferior in terms of PSLR and are more vulnerable to motion‑induced decorrelation; and (iii) a carefully crafted fixed‑point FPGA implementation can deliver the required processing throughput while minimizing silicon area and power, making the proposed ISAC solution attractive for deployment in next‑generation vehicular platforms.