ECASS: Edge computing based auxiliary sensing system for self-driving vehicles

Abstract Self-driving vehicles, combining automobiles with autopilot systems, enable intelligent and safe driving. Self-driving vehicles can achieve accurate automatic navigation, trajectory tracking, and automatic overtaking by using GPS, radars, and inertial measurement unit (IMU). Among them, overtaking is essential in order to avoid excessive waiting time and improve the traffic efficiency. When following a large truck or bus, the self-driving vehicle cannot ensure the safe overtaking because the line-of-sight (LOS) range detected by the radar and camera is blocked, thus unable to perceive the surrounding environment accurately. A commonly adopted mitigation is to follow the truck or bus at a reduced speed, at the cost of reduced traffic efficiency and more traffic jams. To mitigate this deficiency, this paper develops an auxiliary sensing system using edge computing to locate nearby vehicles for self-driving vehicles, called ECASS. Specially, infrastructure deployed along the road like servers are utilized to accurately locate vehicles according to GPS and wireless information such as WiFi or DSRC. Subsequently, the server will transmit the localization information of nearby vehicles to the self-driving vehicle, based on which it can determine the driving state for the next moment despite of the obstruction. Extensive simulations verify that ECASS based trajectory is much closer to the real trajectory than GPS. Especially when GPS error is set within 10 m, ECASS can reduce the mean absolute localization error from more than 7 m to about 3 m.