Risk-Aware Active Inverse Reinforcement Learning

📖 Description

Risk-Aware Active Inverse Reinforcement Learning Daniel S. Brown, Yuchen Cui, and Scott Niekum Department of Computer Science University of Texas at Austin, United States [email removed], [email removed], [email removed] Abstract: Active learning from demonstration allows a robot to query a human for speci?c types of input to achieve ef?cient learning. Existing work has explored a variety of active query strategies; however, to our knowledge, none of these strate- gies directly minimize the performance risk of the policy the robot is learning. Utilizing recent advances in performance bounds for inverse reinforcement learn- ing, we propose a risk-aware active inverse reinforcement learning algorithm that focuses active queries on areas of the state space with the potential for large gen- eralization error. We show that risk-aware active learning outperforms standard active IRL approaches on gridworld, simulated driving, and table setting tasks, while also providing...

📊 Game Impacts

💭 Reactions

🔗 Sources

• [1] https://arxiv.org/abs/1901.02161