Safe MDP planning by learning temporal patterns of undesirable trajectories and averting negative side effects

• Main Authors: LOW, Siow Meng, KUMAR, Akshat, SANNER, Scott
• 格式: text
• 語言: English
• 出版: Institutional Knowledge at Singapore Management University 2023
• 主題: Artificial intelligence; Cost functions; Lagrange multipliers; Learning systems; Artificial Intelligence and Robotics; Databases and Information Systems; Programming Languages and Compilers
• 在線閱讀: https://ink.library.smu.edu.sg/sis_research/8604; https://ink.library.smu.edu.sg/context/sis_research/article/9607/viewcontent/2304.03081.pdf
• 機構: Singapore Management University

In safe MDP planning, a cost function based on the current state and action is often used to specify safety aspects. In real world, often the state representation used may lack sufficient fidelity to specify such safety constraints. Operating based on an incomplete model can often produce unintended negative side effects (NSEs). To address these challenges, first, we associate safety signals with state-action trajectories (rather than just immediate state-action). This makes our safety model highly general. We also assume categorical safety labels are given for different trajectories, rather than a numerical cost function, which is harder to specify by the problem designer. We then employ a supervised learning model to learn such non-Markovian safety patterns. Second, we develop a Lagrange multiplier method, which incorporates the safety model and the underlying MDP model in a single computation graph to facilitate agent learning of safe behaviors. Finally, our empirical results on a variety of discrete and continuous domains show that this approach can satisfy complex non-Markovian safety constraints while optimizing agent's total returns, is highly scalable, and is also better than previous best approach for Markovian NSEs.