Sensing Hidden Vehicles Based on Asynchronous V2V Transmission: A Multi-Path-Geometry Approach
share
Accurate vehicular sensing is a basic and important operation in autonomous driving for autonomous vehicles. The state-of-the-art sensing technologies (such as RADAR and LIDAR) are incapable of detecting hidden vehicles (HVs) without line-of-sight. This is arguably the reason behind some recent fatal accidents involving autonomous vehicles. To address this issue, this paper presents novel HV-sensing technologies that exploit multi-path transmission from a HV to a sensing vehicle (SV). The powerful technology enables the SV to detect multiple HV-state parameters including position, orientation of driving direction, and size (approximated by the arrays size). Its implementation does not even require transmitter-receiver synchronization like the conventional microwave positioning techniques. Our design approach leverages estimated information on multi-path (namely their angles-of-arrival, angles-of-departure and time-of-arrival) and their geometric relations. As a result, a complex system of equations or optimization problems, where the desired HVstate parameters are variables, can be formulated for the cases of negligible and non-negligible noise, respectively. The development of intelligent solution methods ranging from least-square estimator to disk and box minimizations yields a set of practical HV-sensing techniques. We study their feasibility conditions in terms of the required number of paths, which are found to be 4-6 depending on the specific solution method. Furthermore, practical solutions, including sequential path combining and random directional beamforming, are proposed to enable the HV-sensing to cope with the scenario of insufficient received paths. Last, realistic simulation of driving in both the highway and rural scenarios demonstrates the effectiveness of the proposed HV-sensing technologies.
READ FULL TEXT