Student Project 2011 - Interactive Reinforcement Learning

Interactive reinforcement learning is reinforcement learning where the agent receives its rewards from a human teacher. The teacher may also give guidance to the agent in order to focus its attention on a specific object of location in the environment. Although it has been introduced a few years ago in a simulated environment, there is no evaluation of the approach in real world. In this paper we studied the importance of human guidance on the performance of learning using a real robot. We also tested the impact of the state space size on learning performance, by teaching the same task in a small and large state space in our real world domain. Our motivation is based on the lack of analysis and comparison of different Learning from Demonstration approaches on a simple common environment in robotics field.

Category:

Science & Technology

Tags:

• robotics
• human robot interaction