RadarTrack: Enhancing Ego-Vehicle Speed Estimation with Single-chip mmWave Radar

TL;DR

RadarTrack addresses robust ego-speed estimation for mobile platforms using a single-chip mmWave radar by adopting a phase-based speed estimation pipeline that operates without cross-modal learning or deep networks. It derives a 4th-order equation in the ego-speed from phase changes, solves for multiple roots per frame, and uses the mode across chirps to select a consistent speed, while separately segmenting static and dynamic reflectors based on radial-speed profiles. The system integrates two main components—static/dynamic object separation and phase-based translational speed computation—implemented on TI IWR1843 hardware with edge processing on a Jetson Nano, achieving a median MAE around $0.02\ \mathrm{m/s}$ and low power, across UGV, UAV, and handheld platforms. The approach delivers real-time, radar-only ego-motion estimation with strong sub-doppler performance and reduced computational load, enabling practical use in micro-robotics, augmented reality, and autonomous navigation without reliance on bulky sensors or DL models.

Abstract

In this work, we introduce RadarTrack, an innovative ego-speed estimation framework utilizing a single-chip millimeter-wave (mmWave) radar to deliver robust speed estimation for mobile platforms. Unlike previous methods that depend on cross-modal learning and computationally intensive Deep Neural Networks (DNNs), RadarTrack utilizes a novel phase-based speed estimation approach. This method effectively overcomes the limitations of conventional ego-speed estimation approaches which rely on doppler measurements and static surrondings. RadarTrack is designed for low-latency operation on embedded platforms, making it suitable for real-time applications where speed and efficiency are critical. Our key contributions include the introduction of a novel phase-based speed estimation technique solely based on signal processing and the implementation of a real-time prototype validated through extensive real-world evaluations. By providing a reliable and lightweight solution for ego-speed estimation, RadarTrack holds significant potential for a wide range of applications, including micro-robotics, augmented reality, and autonomous navigation.

Figures (9)

• Figure 1: Comparison of doppler-based and phase-based speed estimation for a radar-mounted ego-vehicle moving at (a) 2.8 cm/s, and (b) 6 cm/s. The noise floor indicates the minimum achievable speed detection limit.
• Figure 2: System pipeline of RadarTrack.
• Figure 3: (a) For all static points, the radial component (green arrow) of the relative speed which varies sinusoidally, (b) Distribution of ego-vehicle speed estimates, (c) A typical dynamic scene (left) and classification output (right).
• Figure 4: (a) Speeds falling between two doppler bins are assigned to the lower bin due to FFT; resultant of two phasors is shown, (b) A typical scenario where a static object is present in front of a moving radar.
• Figure 5: (a) Different ego-vehicle settings, (b) deployment setup of RadarTrack.