The cooperative braking system, which combines in-wheel motors (IWMs) and electromechanical brakes (EMBs), is a key technology for intelligent chassis systems. However, conventional fault estimation methods are limited by the system's distributed architecture and complex control modes, resulting in compromised accuracy and delayed response. Moreover, tolerant control methods face challenges in adapting to diverse working conditions and ensuring vehicle dynamics stability. Therefore, this study proposes a novel fault estimation method based on a long short-term memory network (LSTM) and an H∞ adaptive observer. This approach leverages the strengths of deep learning and H∞ adaptive algorithms to effectively handle the system's strong nonlinearities and instability. This enables improved fault estimation accuracy and response time in complex scenarios caused by different fault types of brake units. In addition, a fault-tolerant strategy is developed based on the dynamics stability region feedforward method (DSRFM), which considers the current vehicle maneuvering state and brake unit fault status. The control allocation problem's objective function and constraints are dynamically adjusted to optimize brake torque distribution, enhancing the system's adaptability and maneuvering stability for brake unit faults under different dynamic stability regions, while also considering brake energy recovery. Finally, simulation and bench test results validate the effectiveness and superiority of the proposed methods in fault estimation and tolerant control.
Chang et al. (Sat,) studied this question.