A Safety Aware Model-Based Reinforcement Learning Framework for Systems with Uncertainties

Safety awareness is critical in reinforcement learning when task restarts are not available and/or when the system is safety-critical. Safety requirements are often expressed in terms of state and/or control constraints. In the past, model-based reinforcement learning approaches combined with barrier transformations have been used as an effective tool to learn the optimal control policy under state constraints for systems with fully known models. In this paper, a reinforcement learning technique is developed that utilizes a novel filtered concurrent learning method to realize simultaneous learning and control in the presence of model uncertainties for safety-critical systems.