Automatic emergency obstacle avoidance for intelligent vehicles considering driver-environment risk evaluation

Abstract

Obstacle avoidance is crucial for driving safety, especially in curve road scenarios. To improve the driving safety, this paper proposes an automatic emergency obstacle avoidance strategy for intelligent vehicles with integrated consideration of the driver risk and environment risk evaluation. First, a framework for driver risk evaluation based on distraction detection and driver body pose estimation is proposed. Driver risk is obtained by fusing the pose deviation level obtained by BlazePose and the distraction type detected based on Swin Transformer. Then, an adaptive driving risk evaluation model based on Gaussian model is established by analyzing the characteristics of curve road, which can accurately describe the curve road risk distribution. Subsequently, an automatic emergency obstacle avoidance strategy integrating driver-environment risk is established based on the human-machine cooperative driving pattern and game theory. The cooperative path planning provides safe obstacle avoidance paths. Finally, driver-in-the-loop experiments are conducted to validate the effectiveness of the proposed strategy in curve road scenarios. The results demonstrate that the proposed strategy has superior performance than other advanced cooperative driving strategy in improving driving safety, reducing tracking error, and enhancing vehicle stability and driving comfort, etc.