Abstract: System safety is the primary premise of autonomous driving related research.The local path planning algorithm makes path planning based on local environmental information to ensure that vehicles can successfully pass through specific areas to avoid obstacles and ensure the safety of autonomous vehicles.In order to solve the problem of curve path planning failure caused by the limited visual range of the camera,based on the original Delaunay triangulation path planning algorithm,unilateral cone bucket translation and curve fitting are added to the curve area to realize local path planning,so as to ensure the driving safety of unmanned vehicles.The experimental results show that the planning success rate in the curve driving process is increased from 36.6% to 92.4%,and the average time of a single pathing is 0.264 ms,which is 13.7% less than the original algorithm.The improved Delaunay triangulation algorithm can improve the path planning success rate when the number of unilateral cone barrels is insufficient.In addition,it improves the timeliness of path planning to a certain extent and ensures the safe driving of unmanned vehicles on curves.