Automotive Radar Research Articles

Mutual Interference Avoidance of FMCW Automotive Radars Using AI

Abstract. This study introduces a novel frequency-selective radar system for street condition monitoring (SCM) that eliminates the cost-intensive dedicated monolithic microwave integrated circuit (MMIC) channels approach by sharing radar channels across different frequency bands. The proposed design uses a step-impedance filter (SIF) to allow the radar to perform standard operations at 76.5 GHz while adding SCM capabilities at 80.5 GHz. The radar system ensures robust frequency selectivity, meeting strict isolation requirements to minimize mutual coupling between the antennas operating at different frequencies. An intensive investigation is conducted to determine the minimum co-polar to cross-polar isolation required for effective frequency-selectivity in the system. Simulations show how the SCM antenna affects the primary radar antenna at 76.5 GHz and how the primary radar antenna affects the SCM antenna at 80.5 GHz. The results confirm that this design can integrate SCM without affecting the main radar's performance, offering a practical and cost-effective solution for autonomous vehicles.

Automotive radar interference suppression using improved dual-path convolutional recurrent network

Direct Multi-Target Localization With Cooperative Automotive Radar

Automotive radar is known as a promising sensing technology in autonomous vehicles due to its reliability. In current autonomous vehicles, the 77 – 81 GHz frequency band is the principal operating band for automotive radars. For the efficient operation of automotive radars, the radar antenna needs to be highly accurate. However, higher operating frequencies may present challenges in radar antenna design, requiring high gain, wide bandwidth, and low sidelobe levels (SLL). To address this issue, this study aims to adapt a planar series-fed linear antenna array to 79 GHz automotive radar applications using three different grounded coplanar waveguide (GCPW) feed configurations, including coplanar gap source port, vertical ground bridge, and wave port. Simulations were conducted to evaluate the performance of the antenna with the feed configurations. According to the results, it was shown that the antenna with wave port feed achieved the best impedance bandwidth (>3 GHz), whereas the antenna with either the coplanar gap source port or the vertical ground bridge configurations exhibited better main lobe phase centering and a higher gain (>18.4 dBi), with an acceptable SLL below -16.28 dB. It is believed that these findings may contribute to the development of high-performance radar antennas for next-generation autonomous vehicles.

This paper proposes and evaluates two neural network-based approaches for object classification in automotive radar systems, comparing the performance impact of grid search and genetic algorithm (GA) hyperparameter optimization strategies. The task involves classifying cars, pedestrians, and cyclists using radar-derived features. The grid search–optimized model employs a compact architecture with two hidden layers and 10 neurons per layer, leveraging kinematic correlations and motion descriptors to achieve mean accuracies of 90.06% (validation) and 90.00% (test). In contrast, the GA-optimized model adopts a deeper architecture with nine hidden layers and 30 neurons per layer, integrating an expanded feature set that includes object dimensions, signal-to-noise ratio (SNR), radar cross-section (RCS), and Kalman filter–based motion descriptors, resulting in substantially higher performance at approximately 97.40% mean accuracy on both validation and test datasets. Principal Component Analysis (PCA) and SHapley Additive exPlanations (SHAP) highlight the enhanced discriminative power of the new set of features, while parallelized GA execution enables efficient exploration of a broader hyperparameter space. Although currently optimized for urban traffic scenarios, the proposed approach can be extended to highway and extra-urban environments through targeted dataset expansion and developing additional features that are less sensitive to object kinematics, thereby improving robustness across diverse motion patterns and operational contexts.

Automotive Radar Multi-Frame Track-Before-Detect Algorithm Considering Self-Positioning Errors

In response to the urgent need for ranging resolution, real-time target detection, and point cloud clustering in high-precision 4D millimeter-wave radar imaging, this paper systematically reviews the research progress of key signal processing algorithms. Regarding ranging accuracy, while spectrum refinement algorithms (such as ZFFT and CZT) and super-resolution algorithms (such as MUSIC and compressed sensing) improve resolution, they generally suffer from high computational complexity and insufficient utilization of phase information. In the field of constant false alarm rate (CFAR), mean-value algorithms (CA-CFAR) offer excellent real-time performance but weak multi-target detection capabilities, while ordered statistics algorithms (OS-CFAR) offer strong interference tolerance but require optimization of the adaptive range threshold. Among clustering algorithms, the improved 3D PG-DBSCAN overcomes the global density limitations of traditional DBSCAN, but its static grid parameter setting still restricts its adaptability to dynamic scenes. Based on this, this paper proposes a coherent information-fused CZT-Rife joint ranging algorithm, a range-adaptive ED-CFAR detection strategy, and a 3D PG-DBSCAN optimization scheme with dynamic grid parameters, providing theoretical support for high-precision real-time processing in automotive millimeter-wave radars.

This paper presents the design and development of a compact multi-band high-frequency antenna tailored for millimeter-wave applications, particularly within the 6G frequency range. The antenna features a small footprint of 10 × 12 × 1.5 mm2 and is designed using advanced electromagnetic simulations in CST Microwave Studio. Fabrication was carried out on an FR4 substrate, selected for its favorable properties at high frequencies. The antenna demonstrates an exceptionally wide impedance bandwidth of approximately 166%, covering a broad frequency range from 9.1 GHz to 100 GHz with a central frequency near 45.45 GHz. It exhibits stable radiation characteristics across the operating band, achieving a peak gain of 7.95dBi and an overall efficiency of 85%. Its miniaturized form factor and broad operational range make it a strong candidate for a wide spectrum of applications, including X-band radar, Ku-band satellite communications, K-band sensing, Ka-band 5G systems, V-band short-range wireless, W-band automotive radar, and future technologies such as 6G and security imaging systems.

Abstract. There is increasing interest in automotive sensor monitoring systems as a means to enhance safety by providing reliable assistance in hazardous situations. These systems are commonly based on video cameras; however, their effectiveness is significantly reduced in adverse weather conditions such as fog, rain, or in the presence of smoke. To address this limitation, radar sensors—particularly imaging radars—are gaining prominence within the context of Driver Assistance Systems. A key challenge in current radar signal processing techniques is their limited ability to distinguish multiple targets along the same line of sight. In this paper, we propose a novel radar signal processing approach based on Deep Learning, capable of detecting and differentiating two or more targets aligned on the same line of sight, while also estimating the position and speed of vehicles ahead. Specifically, we adapt techniques originally developed for civil and military tracking radar applications to the automotive context, taking into account the higher spatial resolution and lower signal-to-noise ratio (SNR) characteristic of automotive radars. The proposed system integrates target detection, tracking, recognition, classification, and analysis, with a particular focus on the accurate identification of close-range targets.

Traditional 3D millimeter-wave radars lack target height information, leading to identification failures in complex scenarios. Upgrading to 4D millimeter-wave radars enables four-dimensional information perception, enhancing obstacle detection and improving the safety of autonomous driving. Given the high cost-sensitivity of in-vehicle radar systems, single-chip 4D millimeter-wave radar solutions, despite technical challenges in imaging, are of great research value. This study focuses on developing single-chip 4D automotive millimeter-wave radar, covering system architecture design, antenna optimization, signal processing algorithm creation, and performance validation. The maximum measurement error is approximately ±0.2° for azimuth angles within the range of ±30° and around ±0.4° for elevation angles within the range of ±13°. Extensive road testing has demonstrated that the designed radar is capable of reliably measuring dynamic targets such as vehicles, pedestrians, and bicycles, while also accurately detecting static infrastructure like overpasses and traffic signs.

Fabrication of NbMoTaW High-Entropy-Alloy Thin Films for Application in the Monopole Antennas of Automotive Radar

Holographic metasurface antennas (HMAs) have gained significant attention for their ability to achieve dynamic beam steering, high gain, and compact integration in high-frequency communication systems. This paper presents the design, analysis, and fabrication of a holographic metasurface antenna operating at the 24 GHz frequency band, targeting applications such as automotive radar, wireless communications, and satellite links. The proposed antenna employs a metasurface layer to manipulate electromagnetic wavefronts based on holographic principles, enabling efficient beamforming and reduced sidelobe levels. Full-wave electromagnetic simulations are conducted to optimize the metasurface structure, ensuring enhanced radiation efficiency and wide-angle beam scanning. A prototype is fabricated using PCB-based manufacturing techniques, and its performance is experimentally validated. Measurement results demonstrate a high-gain radiation pattern, effective beam steering capabilities, and improved efficiency, making the proposed HMA a promising candidate for next-generation high-frequency communication systems.

Deep generative models (DGMs) have been studied and developed primarily in the context of natural images and computer vision. This has spurred the development of (Bayesian) methods that use these generative models for inverse problems in image restoration, such as denoising, inpainting and super-resolution. In recent years, generative modelling for Bayesian inference on sensory data has also gained traction. Nevertheless, the direct application of generative modelling techniques initially designed for natural images on raw sensory data is not straightforward, requiring solutions that deal with high dynamic range signals (HDR) acquired from multiple sensors or arrays of sensors that interfere with each other, and that typically acquire data at a very high rate. Moreover, the exact physical data-generating process is often complex or unknown. As a consequence, approximate models are used, resulting in discrepancies between model predictions and observations that are non-Gaussian, in turn complicating the Bayesian inverse problem. Finally, sensor data are often used in real-time processing or decision-making systems, imposing stringent requirements on, e.g. latency and throughput. In this article, we discuss some of these challenges and offer approaches to address them, all in the context of high-rate real-time sensing applications in automotive radar and medical imaging.This article is part of the theme issue ‘Generative modelling meets Bayesian inference: a new paradigm for inverse problems’.

Emerging Trends in Radar: Automotive Radar Networks

Mono RGB cameras and automotive radar sensors provide a complementary information set that makes them excellent candidates for sensor data fusion to obtain robust traffic user detection. This has been widely used in the vehicle domain and recently introduced in roadside-mounted smart infrastructure-based road user detection. However, the performance of the most commonly used late fusion methods often degrades when the camera fails to detect road users in adverse environmental conditions. The solution is to fuse the data using deep neural networks at the early stage of the fusion pipeline to use the complete data provided by both sensors. Research has been carried out in this area, but is limited to vehicle-based sensor setups. Hence, this work proposes a novel deep neural network to jointly fuse RGB mono-camera images and 3D automotive radar point cloud data to obtain enhanced traffic user detection for the roadside-mounted smart infrastructure setup. Projected radar points are first used to generate anchors in image regions with a high likelihood of road users, including areas not visible to the camera. These anchors guide the prediction of 2D bounding boxes, object categories, and confidence scores. Valid detections are then used to segment radar points by instance, and the results are post-processed to produce final road user detections in the ground plane. The trained model is evaluated for different light and weather conditions using ground truth data from a lidar sensor. It provides a precision of 92%, recall of 78%, and F1-score of 85%. The proposed deep fusion methodology has 33%, 6%, and 21% absolute improvement in precision, recall, and F1-score, respectively, compared to object-level spatial fusion output.

This paper proposes a compact 2 × 2 on-chip microstrip antenna array operating for W-band applications. The design utilizes a low-loss glass substrate to mitigate dielectric losses and integrates an embedded feeding structure with wideband T-junction power dividers, addressing bandwidth limitations and feed network losses in conventional approaches. Experimental results demonstrate a relative bandwidth of 10.1% (76.11–83.87 GHz) with gain exceeding 10 dBi across the bandwidth, closely aligning with simulated predictions. This work provides a cost-effective solution for millimeter-wave and terahertz antenna systems, balancing high-performance requirements with fabrication simplicity for automotive radar and 5G/6G communication applications.

Millimeter-wave antennas have become more important recently due to the diversity of applications in 5G and upcoming 6G technologies, of which automotive systems constitute a significant part. Two crucial indices, detection range and angular resolution, are used to distinguish the performance of the automotive antenna. Strong gains and narrow beamwidths of highly directive radiation beams afford longer detection range and finer spatial selectivity. Although conventionally used, patch antennas suffer from intrinsic path losses that are much higher when compared to the waveguide antenna. Designed at 77 GHz, presented in this article is an 8-element slot array on the narrow side wall of a rectangular waveguide, thus being readily extendable to planar arrays by adding others alongside while maintaining the element spacing requirement for grating lobe avoidance. Comprising tilted Z-shaped slots for higher gain while keeping constrained within the narrow wall, adjacent ones separated by half the guided wavelength are inclined with reversed tilt angles for cross-polar cancelation. An open-ended external waveguide is placed over each slot for polarization purification. Equivalent circuit models of slotted waveguides aid the design. An approach for sidelobe suppression using the Chebyshev distribution is adopted. Four types of arrays are proposed, all of which show potential for different demands and applications in automotive radar. Prototypes based on designs by simulations were fabricated and measured.

Background: Modern autonomous driving systems impose high demands on the quality of object detection and classification in the surrounding environment. Radar systems, due to their resilience to adverse weather conditions and ability to measure velocity, play a crucial role among the object and obstacle detection systems used in autonomous vehicles. However, various technical issues related to noise, incorrect classification, and errors in determining object characteristics can hinder the operation of these systems. Objective: Identification and analysis of the main problems of object detection and classification based on radar data, and assessment of their impact on the safety and performance of autonomous driving systems. Methods: In this study, experimental data collection was carried out in city traffic conditions using the ARS 408 automotive radar. Modern software tools including the Robot Operating System (ROS) were used to analyze and process the data. Detection quality evaluation metrics such as IoU, Precision, Recall and F1-score were applied in the study. Results: Within the study, the methodology for radar system data analysis and identification of the main problems encountered during object detection, including the effects of noise, classification errors and object size biases, is developed. Approaches to assessment of quality of the detection algorithms are proposed and a comparative analysis of the convergence of object detection data under various conditions is carried out. Conclusions: The results highlight the main problems of object detection by radar systems and help to assess the quality of current algorithms. The practical significance of the study lies in analyzing the weaknesses of object detection systems and providing data for algorithm improvement, which can enhance the safety of autonomous vehicles.

Abstract Automotive radar systems face the challenge of managing high sampling rates and large data bandwidth while complying with stringent real-time and energy efficiency requirements. Neuromorphic computing offers promising solutions because of its inherent energy efficiency and parallel processing capacity. Yet, most sensor systems, such as radars, do not produce suitable data for further neuromorphic processing of the data. This research presents a novel spiking neuron model for signal processing of frequency-modulated continuous wave (FMCW) radars that outperforms the state-of-the-art spectrum analysis algorithms in latency and data bandwidth. These spiking neural resonators are based on the resonate-and-fire neuron model and optimized to dynamically process raw radar data while simultaneously emitting an output in the form of spikes. We designed the first neuromorphic neural network consisting of these spiking neural resonators that estimates range and angle from FMCW radar data, evaluated the network on simulated automotive datasets and compared the results with a state-of-the-art pipeline for radar processing. The proposed neuron model significantly reduces the processing latency compared to traditional frequency analysis algorithms, such as the Fourier transformation (FT), which needs to sample and store entire data frames before processing. The evaluations demonstrate that these spiking neural resonators achieve state-of-the-art detection accuracy while emitting spikes simultaneously to processing and transmitting only 0.02% of the data compared to a float-32 FT. The results showcase the potential for neuromorphic signal processing for FMCW radar systems and pave the way for designing neuromorphic radar sensors.