Assured learning‐enabled autonomy: A metacognitive reinforcement learning framework

Reinforcement learning (RL) agents with pre-specified reward functions cannot provide guaranteed safety across variety of circumstances that an uncertain system might encounter. To guarantee performance while assuring satisfaction constraints circumstances, assured autonomous control framework is presented in this article by empowering RL algorithms metacognitive capabilities. More specifically, adapting the function parameters agent performed a decision-making layer to assure feasibility agent. That is, learned policy satisfies specified signal temporal logic achieving as much possible. The monitors any possible future violation under actions and employs higher-layer Bayesian algorithm proactively adapt for lower-layer minimize intervention, fitness leveraged metric evaluate success liveness specifications, intervenes only if there risk failure. Finally, simulation example provided validate effectiveness proposed approach.