Autonomous navigation for reconfigurable snake-like robots in challenging, unknown environments

We present a navigation approach for reconfigurable snake-like robots to autonomously overcome unknown and challenging obstacles like stairs or large steps. To calculate convenient motions we use a planning algorithm that takes different optimization criteria like distance, time or energy into account. In addition, the current robot configuration, i.e. the combination of robot modules that make up the complete system, is considered while planning, whereby the planning algorithm can determine whether an obstacle can be overcome or not. To handle constraints like inholonomity of the robot we use a purely geometric planner together with an extension, the Secondary Nearest Neighbor Space, to increase the efficiency of the planning algorithm. Combining multi-stage navigation and motion planning with a follow-the-leader control method keeps the planning time unaffected from a changing amount of robot modules. The developed and implemented methods are presented and three different scenarios have been chosen for evaluation on a real robot: the reconfigurable snake-like robot KAIRO 3. The results show that our approach is very well-suited to distinguish different robot configurations and enables snake-like robots to overcome previously unknown obstacles fully autonomously.