MPC-Inspired Reinforcement Learning for Verifiable Model-Free Control
This addresses the need for verifiable and efficient controllers in robotics and demanding control tasks, though it is incremental as it builds on existing MPC and DRL methods.
The paper tackles the problem of designing verifiable and high-performance model-free controllers by introducing a parameterized controller inspired by Model Predictive Control (MPC) and trained with Deep Reinforcement Learning (DRL), resulting in controllers that match MPC and MLP controllers in control performance while offering superior robustness, computational efficiency, and fewer parameters.
In this paper, we introduce a new class of parameterized controllers, drawing inspiration from Model Predictive Control (MPC). The controller resembles a Quadratic Programming (QP) solver of a linear MPC problem, with the parameters of the controller being trained via Deep Reinforcement Learning (DRL) rather than derived from system models. This approach addresses the limitations of common controllers with Multi-Layer Perceptron (MLP) or other general neural network architecture used in DRL, in terms of verifiability and performance guarantees, and the learned controllers possess verifiable properties like persistent feasibility and asymptotic stability akin to MPC. On the other hand, numerical examples illustrate that the proposed controller empirically matches MPC and MLP controllers in terms of control performance and has superior robustness against modeling uncertainty and noises. Furthermore, the proposed controller is significantly more computationally efficient compared to MPC and requires fewer parameters to learn than MLP controllers. Real-world experiments on vehicle drift maneuvering task demonstrate the potential of these controllers for robotics and other demanding control tasks.