Reinforcement Learning for Mixed-Integer Problems Based on MPC
This work addresses a domain-specific challenge in control systems for applications requiring mixed-integer inputs, representing an incremental improvement over existing actor-critic methods.
The paper tackles the problem of applying reinforcement learning to systems with both continuous and integer inputs by proposing a mixed-integer MPC-based policy approximation, resulting in a computationally inexpensive exploration technique that ensures constraint satisfaction and a compatible advantage function for gradient computation.
Model Predictive Control has been recently proposed as policy approximation for Reinforcement Learning, offering a path towards safe and explainable Reinforcement Learning. This approach has been investigated for Q-learning and actor-critic methods, both in the context of nominal Economic MPC and Robust (N)MPC, showing very promising results. In that context, actor-critic methods seem to be the most reliable approach. Many applications include a mixture of continuous and integer inputs, for which the classical actor-critic methods need to be adapted. In this paper, we present a policy approximation based on mixed-integer MPC schemes, and propose a computationally inexpensive technique to generate exploration in the mixed-integer input space that ensures a satisfaction of the constraints. We then propose a simple compatible advantage function approximation for the proposed policy, that allows one to build the gradient of the mixed-integer MPC-based policy.