Neural Network Based Model Predictive Control for an Autonomous Vehicle
This work addresses the need for verifiable and efficient controllers in autonomous driving, though it appears incremental as it builds on existing learning-based control methods.
The paper tackles the problem of replacing model predictive controllers with learning-based controllers for autonomous vehicles, focusing on the bicycle model and comparing supervised and reinforcement learning approaches to achieve small, efficient neural networks suitable for real-time embedded systems and formal verification.
We study learning based controllers as a replacement for model predictive controllers (MPC) for the control of autonomous vehicles. We concentrate for the experiments on the simple yet representative bicycle model. We compare training by supervised learning and by reinforcement learning. We also discuss the neural net architectures so as to obtain small nets with the best performances. This work aims at producing controllers that can both be embedded on real-time platforms and amenable to verification by formal methods techniques.