Over the past decades, there has been significant research effort dedicated to the development of intelligent vehicles and V2X systems. This paper proposes a road traffic risk assessment method for road traffic accident prevention of intelligent vehicles. This method is based on HMM (Hidden Markov Model) and is applied to the prediction of steering angle status to (1) evaluate the probabilities of the steering angle in each independent interval and (2) calculate the road traffic risk in different analysis regions. According to the model, the road traffic risk is quantified and presented directly in a visual form by the time-varying risk map, to ensure the accuracy of assessment and prediction. Experiment results are presented, and the results show the effectiveness of the assessment strategies.

Brake system failure may occur because of design defects, poor maintenance, or improper operation and is one of the causes of serious traffic accidents. For high-speed vehicles, brake system failure can cause them to almost lose the ability to avoid collisions. Currently, there are no active safety functions for vehicles with near-total brake system failure. In this article, an adaptive drift-control method is proposed to reduce longitudinal speed and avoid collisions after brake system failure. Collision risk is analyzed by combining longitudinal obstacles and lateral traffic environments. The controller adaptively adopts a control policy according to the collision risk. Vehicle modeling is performed considering four-dimensional nonlinear and coupled dynamics in drift control. The control policy is based on a multilayer neural network that is trained using the jump-start soft actor critic algorithm. Initiated from an expert-guided policy, this algorithm iteratively updates the policy. Finally, we test and validate the proposed drift controller in real vehicle tests. The results demonstrate that the adaptive drift controller has the ability to rapidly stop the controlled vehicle and avoid collisions after near-total brake failure. In addition, the lateral displacement during drift is effectively suppressed in real tests, thus avoiding potential lateral collisions and achieving omnidirectional obstacle avoidance.