Stable and Real-Time Extended Object Tracking for Autonomous Driving

Abstract

Object tracking is important for autonomous driving. Extended Object Tracking (EOT) is the method that simultaneously estimates both object motion state and shape, and is more robust than tracking object as a mass point. In order to apply EOT to autonomous driving on actual public roads, we should consider its specific problems. For example, considering object shape changes caused by changes in the visible range and tracking many objects in real time are necessary. In this paper, we propose stable and real-time EOT that can be applied to actual public roads, and verified usefulness of the proposed method.