Safe chance constrained reinforcement learning for batch process control

Reinforcement Learning (RL) controllers have generated excitement within the control community. The primary advantage of RL relative to existing methods is their ability optimize uncertain systems independently explicit assumption process uncertainty. Recent focus on engineering applications has been directed towards development safe controllers. Previous works proposed approaches account for constraint satisfaction through tightening from domain stochastic model predictive control. Here, we extend these plant-model mismatch. Specifically, propose a data-driven approach that utilizes Gaussian processes offline simulation and use associated posterior uncertainty prediction joint chance constraints method benchmarked against nonlinear via case studies. results demonstrate methodology uncertainty, enabling even in presence