Abstract: In order to optimize the length and number of turns of intelligent vehicle planning paths in urban roads, this paper proposes an improved adaptive ant colony algorithm. The raster method is used to grid the urban road environment, the shortest path and the least number of turns are used as the optimization objective function, the neighborhood search range is improved, the 8 neighborhood is expanded to 24 neighborhoods, the initial pheromone is updated using a heuristic function, the pheromone volatility coefficient is dynamically adjusted, and a deadlock strategy of returning to the previous step is proposed, and finally the curvature limitation is considered and the three-time B-spline curve method is used to process smoothing optimization. This study shows that the algorithm reduces the path length by 3.89% and 8.38%, the path nodes by 28% and 24.2%, and the number of convergence by 37.5% and 20%, respectively, compared with the standard ACO and the adaptive algorithm under the randomly generated urban environment map model. The results of this study provide a good theoretical basis for the optimization of intelligent vehicle paths on urban roads.