Action module planning and its application to an experimental climbing robot

This paper presents the application of an action module planning method to an experimental climbing robot named LIBRA. The method searches for a sequence of physically realizable actions, called action modules, to produce a plan for a given task. The search is performed with a hierarchical selection process that uses task and configuration filters to reduce the action module inventory to a reasonable search space. Then, a genetic algorithm search finds a sequence of actions that allows the robot to complete the task without violating any physical constraints. The results for the LIBRA climbing robot show the method is able to produce effective plans.