Holistic adaptive energy-efficient MPC architecture for multi-objective control in over-actuated autonomous vehicles

Abstract

This paper presents two novel MPC architectures for autonomous over-actuated in-wheel vehicles, targeting enhanced energy efficiency, stability, and control performance. The first architecture introduces a hierarchical centralized MPC framework that utilizes a minimal-order model to integrate path-tracking, speed control, and stability control objectives. The second architecture extends the hierarchical framework into a holistic MPC design, incorporating direct energy-efficient torque allocation and a tire stability criterion. Energy efficiency is significantly improved by minimizing total power consumption and enforcing operational constraints to maximize motor efficiency. Central to both architectures is a novel multi-criteria adaptive weighting mechanism that dynamically reconciles conflicting objectives by adjusting control priorities based on real-time error magnitudes and driving conditions. This mechanism not only resolves potential conflicts between objectives but also enhances robustness to modeling inaccuracies, uncertainties, disturbances, and variations in road adhesion, while significantly improving control performance. Validation is conducted through joint simulations in Simulink/Matlab and the SCANeR Studio vehicle dynamics simulator. The findings demonstrate that both architectures achieve substantial energy savings while maintaining computational efficiency, with improved stability, comfort, and precision in path-tracking and speed control under challenging conditions, including high-speed, high-curvature, and low-adhesion scenarios.