On Intelligent Robot Motion Planning via Learning

摘要: Automatic motion planning is one of the basic modules that are needed to increase robot intelligence and usability. Unfortunately, inherent complexity has rendered traditional search algorithms incapable solving every problem in real time. To circumvent this difficulty, we explore alternative allowing human operators participate process. By having operator teach during difficult episodes, should be able learn improve its own capability gradually reduce reliance on operator. In paper, present such a learning framework which both can cooperate achieve real-time automatic planning. enable deeper understanding terms performance, it as simple algorithm provide theoretical analysis behavior. particular, characterize situations useful, quantitative bounds predict necessary training time maximum achievable speedup