Mobile robot navigation in unknown, unstructured environments such as hilly terrains is often complex due to uncertainties associated with classification of terrain features as traversable or non-traversable and identification of paths between start and goal that pose minimum danger to the robot during navigation. Humans possess an uncanny ability to identify paths under the presence of uncertainties and determine whether a terrain feature is safe to traverse or not. Therefore, it is beneficial if such human knowledge is taken advantage of when available. This paper discusses the use of reinforcement learning techniques for navigation in hilly type terrains with and without the human assistance. When a prior knowledge such as low resolution aerial view about the terrain is available, a human expert can suggest paths that are relatively safe to traverse. A value matrix is developed using those expert paths that could be used by the robot to steer from any start position in the terrain to the goal position. Matlab based simulations have been carried out and the results have been presented for robot navigation in a terrain in the presence of multiple expert suggested paths.
- Dynamic Systems and Control Division
Mobile Robot Navigation in Hilly Terrain Using Reinforcement Learning Techniques
- Views Icon Views
- Share Icon Share
- Search Site
Tennety, S, & Kumar, M. "Mobile Robot Navigation in Hilly Terrain Using Reinforcement Learning Techniques." Proceedings of the ASME 2011 Dynamic Systems and Control Conference and Bath/ASME Symposium on Fluid Power and Motion Control. ASME 2011 Dynamic Systems and Control Conference and Bath/ASME Symposium on Fluid Power and Motion Control, Volume 2. Arlington, Virginia, USA. October 31–November 2, 2011. pp. 81-86. ASME. https://doi.org/10.1115/DSCC2011-6096
Download citation file: