Safe Control Allocation of Articulated Heavy Vehicles Using Machine Learning
摘要
As articulated heavy vehicles are over-actuated, achieving a safe control allocation is crucial to ensure stability. This study introduces a machine learning model developed to identify unsafe behaviours and modes, such as jack-knifing and trailer swing, enabling the control scheme to prioritize stability. High-fidelity simulations, focusing on high-risk scenarios, generate data for training the machine learning model. This model is integrated into the control scheme to predict safe braking allocations and prevent unsafe vehicle modes during real-time driving scenarios. Initial tests showed promising results regarding prediction accuracy and a safety margin that can be implemented to further ensure that safe vehicle motion is achieved.