A New Wave in Robotics: Survey on Recent mmWave Radar Applications in Robotics

Abstract

We survey the current state of millimeterwave (mmWave) radar applications in robotics with a focus on unique capabilities, and discuss future opportunities based on the state of the art. Frequency Modulated Continuous Wave (FMCW) mmWave radars operating in the 76--81GHz range are an appealing alternative to lidars, cameras and other sensors operating in the near visual spectrum. Radar has been made more widely available in new packaging classes, more convenient for robotics and its longer wavelengths have the ability to bypass visual clutter such as fog, dust, and smoke. We begin by covering radar principles as they relate to robotics. We then review the relevant new research across a broad spectrum of robotics applications beginning with motion estimation, localization, and mapping. We then cover object detection and classification, and then close with an analysis of current datasets and calibration techniques that provide entry points into radar research.

Figures (6)

• Figure 1: VDEs cause significant degradation of both lidar and camera based measurements in autonomous robotics. Camera and lidar measurements in a smoke-filled indoor environment eliminate most identifiable structure and features required for navigation or classification.
• Figure 2: Example robotic platforms with attached radar systems. From left to right: Fast moving Parkour-Car from the University of Colorado--Boulder equipped with a Texas Instruments system-on-chip radar. Boreas data collection vehicle from the University of Toronto equipped with a modern Navtech scanning radar burnett2022boreas. Ground vehicle from Nanyang Technical University using an older Navtech scanning radar adams2007autonomous. Street-crossing robot attached with two Delphi system-on-chip radar radwan2020multimodal.
• Figure 3: Utilizing the FMCW radar, measurements are derived from the chirp signal and the phase of the IF signal. The range measurements are ascertained based on the properties of the received chirp. Both velocity and angle measurements are determined by leveraging the characteristics of the phase difference.
• Figure 4: The two main types of mmWave radar are (i) SoC Radar and (ii) Scanning Radar. For both types, transmit (TX) and receive (RX) antenna send and receive signals from the environment. An IF is generated by mixing the signals and sampled by an ADC. ADC samples can be stored directly. SoC radar stores samples in a datacube (a), where the sample index represents samples along a single chirp from an ADC, the chirp index represent samples along a series of sent chirps, and the TX-RX pair denotes the specific transmit and receive antenna generated a signal. Scanning radar stores samples along a single chirp and by the angle of the antenna at which the chirp was generated (b). These complex IF samples can be furthered processed with SoC radar generating a 3D+1 or 4D Radar Heatmap (c) by applying a FFT. Values are stored by range, azimuth, and elevation in spherical coordinates (3D) with each value containing both intensity and radial velocity (+1). A non-maximal suppression CFAR filter can be applied to this heatmap to generate radar point returns with radial velocity (d). A similar FFT + CFAR processing generates a measure of intensity along each scanned angle presented in a 2D radar image (e).
• Figure 5: Radar presents several types of noise that are unique compared to other sensors. Speckle noise returns are the most common, with ambiguous clutter circled in cyan. Multipath reflections develop where returns bounce off nearby walls or the ground before hitting the antenna, generating reflections of true targets. A series of repeated returns is circled in red. The original image is sampled from the Mulran dataset kim2020mulran.