Kinect based real time obstacle detection for legged robots in complex environments

Legged robots have different terrain traverse capabilities with ground vehicles. They can across steps and grooves, which are obstacles for ground vehicles. Some new challenges are posed for autonomous legged robot navigation in complex environments. In this paper, we present a Kinect based real time obstacle detection algorithm for legged robots. A graph based method is used to label all traversable areas connecting to a starting point. A spiral searching strategy is proposed to find the most ground-like point for graph based traversal. Some time efficient rules are presented to increase the calculation efficiency. Experiments in both indoor and outdoor environment have been done and the results show that our algorithm can be effectively used in autonomous legged robots navigation.