Learning to maintain safety through expert demonstrations in settings with unknown constraints: A Q-learning perspective

Researchers propose SafeQIL, a new Q-learning algorithm that learns safe policies from expert demonstrations in constrained environments where safety constraints are unknown. The approach balances maximizing task rewards while maintaining safety by learning from demonstrated trajectories that successfully complete tasks without violating hidden constraints.

• →SafeQIL algorithm learns safe policies from expert demonstrations without knowing explicit safety constraints.
• →The method uses Q-learning perspective to balance reward maximization with safety maintenance.
• →The algorithm formulates 'promise' of state-action pairs using Q values that incorporate both rewards and safety assessments.
• →SafeQIL outperforms existing inverse constraint reinforcement learning algorithms on challenging benchmark tasks.
• →The approach addresses the critical problem of learning safe AI behavior in environments with unknown safety constraints.