A road adhesion coefficient-aware adaptive path-planning algorithm for vehicle collision avoidance

To address the issues of irrationality and poor adaptability in existing collision avoidance path-planning algorithms, an adaptive vehicle path-planning method incorporating the road adhesion coefficient is proposed. First, a fusion algorithm integrating a convolutional neural network (CNN) and an unscented Kalman filter (UKF) was employed to estimate the road adhesion coefficient. Second, vehicle lane-change stability was analyzed to establish a constraint relationship between the road adhesion coefficient and the maximum permissible path curvature. Subsequently, a cluster of potential collision avoidance paths was generated using a cubic B-spline curve algorithm, from which the optimal path was selected based on a multi-objective optimization function. Finally, collision avoidance tests were conducted using a MATLAB/Carsim co-simulation platform on road surfaces with adhesion coefficients of 0.3, 0.55, 0.85, and a random value to validate the rationality and adaptability of the planned paths. The results demonstrate that the proposed algorithm, which accounts for the road adhesion coefficient, yields more reasonable paths and exhibits greater adaptability than conventional algorithms. This approach significantly enhances vehicle stability and safety during collision avoidance maneuvers across varying road adhesion conditions.