Abstract—Increasing robotic systems autonomy is a major challenge in ensuring both their performance and ease of application in numerous areas of human activity. The present study attempts to combine several artificial intelligence methods to design self-learning control system for the task of mobile robot motion planning in a complex environment. We propose two main components of the control system: a dynamic path planner based on D* algorithm and terrain type prediction subsystem based on classification trees. Efficiency of both of the subsystems grows with time due to knowledge accumulation, leading the robot to a better maneuvering in the environment filled with obstacles.

Index Terms—autonomous mobile robot, motion planning, graph route planning, classification trees, machine learning

Cite: Sergey V. Manko, Valery M. Lokhin, Sekou A. K. Diane, and Alexander S. Panin, "Autonomous Mobile Robot Self-Learning In Motion Planning Problem," International Journal of Mechanical Engineering and Robotics Research, Vol.4, No. 3, pp. 238-241, July 2015. DOI: 10.18178/ijmerr.4.3.238-241