Autonomous Navigation in Dynamic Environments with Reinforcement Learning and Heuristic
Authors:
- 1 December 2010
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 37-42
Abstract
Researchers have created machines which operate autonomously in complex and changing environments. An important problem that has been widely studied is that of autonomous navigation systems, through which attempts have been made to create mechanisms with their own decision making in complex environments. Ideally, an autonomous navigation agent must have an ability to learn while working in its environment. This acquisition of knowledge may be based on a history of actions taken to make decisions without the guidance of a tutor. The performance of the agent may depend on its ability to learn and adapt. This paper presents a reinforcement learning-based method applied to a navigation problem. Using a Q-Learning algorithm, we propose a model to provide autonomous navigation with a policy modified by information from a greedy heuristic. The model aims to improve the performance of the agent with regard to its navigation task.Keywords
This publication has 2 references indexed in Scilit:
- A test problem generator for non-stationary environments2003
- Reinforcement Learning: A SurveyJournal of Artificial Intelligence Research, 1996