Scheme of Autonomous Vehicle Abnormal Behavior Detection Technology Based on Edge Computing

Rongbo Zhang,Kaiyu Cai,Yuzhou Huang

DOI: https://doi.org/10.1145/3409501.3409508

2020-07-03

Abstract:As the development of automatic driving technology, more and more attention has been paid to the safety and standardization of autonomous vehicles. Although high-definition map[4] navigation has been used to assist the automatic driving of vehicles, there are still abnormal behaviors such as illegal retrograde, forced lane change or illegal guidance caused by human intervention driving or runaway behaviors like tire blowout and skidding of vehicles. It's a worthy topic that studying to identify autonomous vehicle abnormal behavior, stop and warn surrounding vehicles to pay attention to avoidance. The existing technology mainly uses the video camera fixed on the traffic light to track the moving vehicle, and forms the vehicle's moving track by tracking the vehicle target, which provides a reliable and scientific data source for vehicles abnormal behavior recognition. However, there will be a high load when all the video flows are gathered in the cloud server for processing, so it is necessary to introduce the edge computing framework to offload [18] the image processing task to edge computing node. Each edge computing node is responsible for monitoring the vehicles within its coverage. If there is no abnormal behavior, the video will not be processed too much. If any abnormal behavior is found, other vehicles in the area will be notified first to avoid the collision or other traffic accidents, and the abnormal information of the vehicles will be uploaded to the cloud server. It can be inferred that the scheme proposed in this paper can make normal driving vehicles react to abnormal behavior vehicles more quickly, correct driving strategies in time, and prevent traffic accidents.