Multi-Vehicle Collision Avoidance by Vehicle Longitudinal Control Based on Optimal Collision Distance Estimation

This paper proposes a collision avoidance method for vehicle longitudinal velocity control based on multi-vehicle collision distance estimation. The method begins by estimating the position and shape of object vehicles with collision risk using environmental sensors. The collision point is identified from the object vehicle's surface, and a Kalman filter is applied for accurate estimation. The optimal collision distance is then determined by evaluating the collision risk at the identified point. A longitudinal control technique, incorporating the optimal collision distance and time gap, is employed to implement the collision avoidance system. The proposed method was validated through scenario-based simulations involving multi-vehicle collision avoidance, which were implemented in an environment combining ROS and the MORAI simulator, along with comparative experiments. Comparative studies with conventional vehicle center-based approaches demonstrated that the proposed surface-based collision point method significantly enhances collision avoidance performance. While the conventional method led to a collision between the ego and object vehicles, the proposed method successfully avoided the collision by maintaining a separation of about 3.6 m, demonstrating its feasibility and reliability.