Radar-Based Localization For Autonomous Ground Vehicles In Suburban Neighborhoods

TL;DR

This paper tackles the challenge of high-precision localization for autonomous ground vehicles in suburban environments where GPS and vision/LiDAR approaches can fail or underperform. It introduces Radar-Inertial Odometry (RIO), a sliding-window optimization that fuses radar Doppler, IMU data, and a novel radar landmark-based heading constraint to produce accurate, high-rate 6DoF pose estimates while robustly excluding dynamic objects. Complementing RIO, the authors propose radar-based mapping with an occupancy grid map and a two-stage radar map matching pipeline to achieve global localization against a pre-built radar map. Experimental results show competitive translation and heading accuracy with lidar-based methods, strong resilience to challenging conditions, and low computational demands suitable for embedded hardware, highlighting the practical viability of radar-centric localization for suburban AGVs.

Abstract

For autonomous ground vehicles (AGVs) deployed in suburban neighborhoods and other human-centric environments the problem of localization remains a fundamental challenge. There are well established methods for localization with GPS, lidar, and cameras. But even in ideal conditions these have limitations. GPS is not always available and is often not accurate enough on its own, visual methods have difficulty coping with appearance changes due to weather and other factors, and lidar methods are prone to defective solutions due to ambiguous scene geometry. Radar on the other hand is not highly susceptible to these problems, owing in part to its longer range. Further, radar is also robust to challenging conditions that interfere with vision and lidar including fog, smoke, rain, and darkness. We present a radar-based localization system that includes a novel method for highly-accurate radar odometry for smooth, high-frequency relative pose estimation and a novel method for radar-based place recognition and relocalization. We present experiments demonstrating our methods' accuracy and reliability, which are comparable with \new{other methods' published results for radar localization and we find outperform a similar method as ours applied to lidar measurements}. Further, we show our methods are lightweight enough to run on common low-power embedded hardware with ample headroom for other autonomy functions.

Figures (6)

• Figure 1: Factor graph used for the RIO estimator, with increasing time going left. The Doppler and Landmark errors $e_d$ and $e_l$ provide constraints on the state that results in a nonlinear minimization problem over these errors by adjusting the state variables $x^i$. The states are connected through binary constraints $e_i$ induced by IMU measurements. A prior is formed from marginalized states prior to $x^0$ which have fallen off of the sliding window, thereby maintaining a constant-time algorithm.
• Figure 2: A neighborhood trajectory and single global map generated using the volumetric mapping method.
• Figure 3: Example of binning raw radar detections into spatial histograms to elucidate the structure of the environment.
• Figure 4: Detailed view of a radar-based OGM showing correspondences to an aerial view of the same area.
• Figure 5: Fields of view of the robot's radar sensors. Above a side view of the robot shows the rear sensor's vertical field of view and below a top-down view shows all three sensors' horizontal fields of view.