🤖 AI Summary
Model predictive control (MPC) for neural state-space models (NSSMs) of vehicle dynamics faces severe computational challenges due to the non-convexity and high dimensionality of the resulting optimization problem. Method: This paper proposes Model Predictive Inference Control (MPIC), a novel paradigm that reformulates motion planning as a Bayesian state estimation task. We design a particle smoothing algorithm based on an ensemble of unscented Kalman filters, achieving high estimation accuracy while improving sampling efficiency and real-time performance. Contribution/Results: Unlike gradient-based MPC methods, MPIC overcomes the exponential growth in computational complexity with neural network scale under learned dynamics. In extensive multi-scenario simulations, MPIC significantly improves both solution efficiency and trajectory quality, enabling, for the first time, efficient closed-loop control using large-scale NSSM-based MPC.
📝 Abstract
Model predictive control (MPC) has proven useful in enabling safe and optimal motion planning for autonomous vehicles. In this paper, we investigate how to achieve MPC-based motion planning when a neural state-space model represents the vehicle dynamics. As the neural state-space model will lead to highly complex, nonlinear and nonconvex optimization landscapes, mainstream gradient-based MPC methods will be computationally too heavy to be a viable solution. In a departure, we propose the idea of model predictive inferential control (MPIC), which seeks to infer the best control decisions from the control objectives and constraints. Following the idea, we convert the MPC problem for motion planning into a Bayesian state estimation problem. Then, we develop a new particle filtering/smoothing approach to perform the estimation. This approach is implemented as banks of unscented Kalman filters/smoothers and offers high sampling efficiency, fast computation, and estimation accuracy. We evaluate the MPIC approach through a simulation study of autonomous driving in different scenarios, along with an exhaustive comparison with gradient-based MPC. The results show that the MPIC approach has considerable computational efficiency, regardless of complex neural network architectures, and shows the capability to solve large-scale MPC problems for neural state-space models.
