An Incremental Learning Approach to Motion Planning with Roadmap Management

Abstract

Traditional approaches to the motion-planning problem can be classified into solutions\r\nfor single-query and multiple-query problems with the tradeoffs on run-time computation\r\ncost and adaptability to environment changes. In this paper, we propose a novel\r\napproach to the problem that can learn incrementally on every planning query and effectively\r\nmanage the learned road-map as the process goes on. This planner is based on previous\r\nwork on probabilistic roadmaps and uses a data structure called Reconfigurable\r\nRandom Forest (RRF), which extends the Rapidly-exploring Random Tree (RRT) structure\r\nproposed in the literature. The planner can account for environmental changes while\r\nkeeping the size of the roadmap small. The planner removes invalid nodes in the roadmap\r\nas the obstacle configurations change. It also uses a tree-pruning algorithm to trim\r\nRRF into a more concise representation. Our experiments show that the resulting roadmap\r\nhas good coverage of freespace as the original one. We have also successful incorporated\r\nthe planner into the application of intelligent navigation control.