Ego-Motion Estimation and Moving Object Tracking using Multi-layer LIDAR

This paper presents an approach for the robust recognition of a complex and dynamic driving environment, such as an urban area, using on-vehicle multi-layer LIDAR. The multi-layer LIDAR alleviates the consequences of occlusion by vertical scanning; it can detect objects with different heights simultaneously, and therefore the influence of occlusion can be curbed. The road environment recognition algorithm proposed in this paper consists of three procedures: ego-motion estimation, construction and updating of a 3-dimensional local grid map, and the detection and tracking of moving objects. The integration of these procedures enables us to estimate ego-motion accurately, along with the positions and states of moving objects, the free area where vehicles and pedestrians can move freely, and the 'unknown' area, which have never previously been observed in a road environment.