Obstacle detection based on a four-layer laser radar

In the case of indoor/urban navigation, obstacles are typically defined as surface points that are higher than the ground plane. However, this characterization cannot be used in cross-country and unstructured environments, where the notion of "ground plane" is often unmeaning. The paper proposes a new obstacle detection algorithm based on a four-layer laser radar (LD_ML) which is applied to an autonomous land vehicle (ALV) in rough terrain. An obstacle is defined by the cluster gradient and height of candidate obstacle points. The obstacle detection algorithm is proposed by analyzing obstacles characterization, which includes four steps: First, obtain the candidate obstacle points through gradient condition; second, collect candidate obstacle points according to the rule of the nearest distance; third, decide a cluster whether it is an obstacle or not according to the cluster height; finally, estimate and predict the position of obstacles. The experiment results testify the algorithm is reliable and stable.