Are We Ready for Radar to Replace Lidar in All-Weather Mapping and Localization?

Abstract

In this work, we demonstrate continuous-time radar-inertial and lidar-inertial odometry using a Gaussian process motion prior. Using a sparse prior, we demonstrate improved computational complexity during preintegration and interpolation. We use a white-noise-on-acceleration motion prior and treat the gyroscope as a direct measurement of the state while preintegrating accelerometer measurements to form relative velocity factors. Our odometry is implemented using slidingwindow batch trajectory estimation. To our knowledge, our work is the first to demonstrate radar-inertial odometry with a spinning mechanical radar using both gyroscope and accelerometer measurements. We improve the performance of our radar odometry by 19% by incorporating an IMU. Our approach is efficient and we demonstrate real-time performance. Code for this project can be found at: https://github.com/utiasASRL/steam icp

Lidar-Level Localization With Radar? The CFEAR Approach to Accurate, Fast, and Robust Large-Scale Radar Odometry in Diverse Environments

“…3. Recent spinning radars without Doppler measurements have demonstrated high range, accuracy, and richness and have inspired numerous methods; from odometry estimation and alignment quality assessment to global localization [10], [14], [40]- [44], localization in previous maps [41] and SLAM [2], [15].…”

“…Burnett et al [9] presented a sliding window batch odometry estimation framework and found that a larger window up to at least 4 scans improved online odometry. Later, Burnett et al [10] used local submaps of radar detections aggregated from 3 scans within a teach and repeat localization framework [10]. Despite this, no systematic evaluation has been carried out to investigate the principled benefit of considering additional scans in spinning radar research.…”

“…In contrast, radar is largely resilient to these conditions [2], [9], [10], and receives increasingly higher attention for various perception tasks [11]. Hence, radar has great potential to enable robust localization in a wide range of scenarios; from autonomous cars and logistics where uninterrupted autonomy is required, to work machines surrounded by fugitive dust when operating on unpaved roads and within underground mines or construction sites.…”

Sensor Fusion in Autonomous Vehicle with Traffic Surveillance Camera System: Detection, Localization, and AI Networking

“…Although in the conventional system the length is calculated by LiDAR 3D for obtaining the 3D bounding box, the proposed model’s 3D formation accuracy will be more accurate because of having an actual length received from the surveillance camera system. Moreover, researchers are working to replace LiDAR [ 61 ] with advanced RADAR systems as well as multi-sensed 3D camera imaging with AIP technologies because of some commercial use limitations of LiDAR such as its high expense, high signal interference and noise, and the problem of having rotating parts. “Will have or have not LiDAR”, the detection by surveillance camera systems, will be supportive for self-driving AVs in all conditions.…”

Fast Local Planning and Mapping in Unknown Off-Road Terrain

“…Although better results are achieved with a fusion of radar and lidar. A more recent study by [5] extends the radar SLAM topic by comparing three localization systems: radar-only, lidaronly, and a cross-modal radar-to-lidar system across varying seasonal and weather conditions. their comparison shows that, with modern algorithms, lidar localization may not preform as poorly in bad weather as other studies have shown.…”