Ground-Vehicle Relative Position Estimation with UWB Ranges and a Vehicle Dynamics Model

Abstract

An accurate relative position vector (RPV) solution between local vehicles is critical in numerous autonomous vehicle applications from cooperative cruise control (CACC) to commercial truck platooning. Relative GPS solutions can deliver sufficient accuracy for such applications, but resiliency and availability can suffer in cluttered environments; thus, relative position solutions independent of GPS are necessary. Ultra-wideband (UWB) radios can provide accurate and robust range, and have gathered attention in recent decades for use in navigation. Prior work has shown that relative position estimation of dynamic vehicles is possible without a GPS reference through UWB-ranging and a constant-velocity Extended Kalman Filter (EKF). Through a kinematic bicycle model, this work modifies the state vector to include direct estimates of the lead vehicle's longitudinal speed and effective steer angle. This improves estimation quality and robustness in the presence of UWB measurement errors, unfavorable relative geometry, and dynamic maneuvers. The algorithm has been compared to previous methods in simulation and experimental testing environments and exhibits both superior accuracy as well as improved robustness during periods of difficult relative geometry.