Enhancing Relative Positioning Accuracy in Vehicle-to-Vehicle Safety: Multivehicle Kinematics and Nonlinear Particle Filtering Approach

Abstract

Unsatisfactory relative positioning accuracy is a bottleneck in collision avoidance applications of vehicle-to-vehicle communication. Previous efforts focused mainly on either enhancing absolute GPS data precision or experimenting with positioning techniques alternative to GPS, such as real-time kinematics, Doppler measurements, or integration of GPS with ultrawideband ranging radios. This paper proposes a new approach that is independent from the sources and precisions of raw data. Based on a multivehicle kinematics model adapted from a multirobot domain, the method can predict vehicles’ distances and bearings by fully utilizing the shared real-time motion data from dedicated short-range communications’ basic safety message packets. Three types of nonlinear filters were proposed and compared to minimize the errors caused by raw data noise and system timing inaccuracies. Experiments indicated that the extended Kalman filter failed to estimate a highly nonlinear process, unscented Kalman filter dysfunctions, if the statistical features of errors were misestimated; only the particle filter maintained satisfactory and stable accuracies in both high-nonlinearity and error-unknown situations. The combination of the kinematics model and particle filtering can serve as an algorithmic complement to various approaches focused on data sources to enhance relative positioning precision.