Data-Driven Robust Preview Control of Vehicle Active Suspension Considering Random Controller Area Network Communication Dropouts

With the shift to full-by-wire chassis architectures, active suspension control is progressively integrated into chassis domain controllers to achieve coordinated chassis management. However, random packet dropouts in controller area network (CAN) communication under high-load conditions can significantly degrade suspension control performance. To address this challenge, this study proposes a novel data-driven robust preview control method. First, the packet-dropout phenomenon in CAN communication is modeled as a Bernoulli random process, and an augmented state-space model of the active suspension system is constructed by incorporating road preview information. Second, based on zero-sum game theory, road disturbances and control inputs are modeled as adversarial players, leading to the formulation of a stochastic game algebraic Riccati equation (SGARE) for the suspension system. To improve data efficiency and reduce design complexity, a data-driven value iteration (VI) reinforcement learning algorithm is employed to approximate the optimal control solution, with rigorous proof of convergence. Simulation results demonstrate that the proposed algorithm provides effective and feasible solutions across different packet-dropout probabilities. Furthermore, hardware-in-the-loop simulations confirm the robustness and reliability of the proposed control scheme, showing that the active suspension system maintains stable performance even in the presence of random CAN communication losses.