DOI: 10.5176/2251-2179_ATAI12.26
Authors: Saleha Raza, Sajjad Haider
Abstract: The paper applies imitation based learning to learn strategies for autonomous agents. Imitation based learning involves learning from an expert by observing him/her demonstrating a task and then replicating it. The effectiveness of the proposed methodology has been assessed in the domain of RoboCup Soccer Simulation 3D, by considering different soccer related tasks to be performed by an autonomous soccer agent. These tasks include (a) dribbling the ball to opponent’s goal and (b) navigating to a target point while avoiding collisions with obstacles. For each task, human demonstrators controlled soccer agents via game controllers and demonstrated them how to perform a specific task. The data gathered during this phase is used as training data to build a classification model which is later used by our soccer agent to perform decision making during matches. Performance of the imitating agent is compared with that of the human-driven agent and the results are very promising as the performances of the two agents match closely.
Keywords: Imitation learning, learning by demonstration, RoboCup Soccer, strategy building
Updating...