The traditional ant colony algorithm in large-scale and complex environment has problems of slow global search convergence, too many turns in the path and not smooth enough. To address these issues, an improved ant colony algorithm is proposed in this article. This method speeds up the convergence of the algorithm by dynamically updating the pheromones on different levels of ant paths. By introducing the distance function and the direction function as heuristic factors, the quality of path search is improved. An improved adaptive pseudo-random transition strategy is utilized to avoid the probability of falling into the local optimal solution. Based on the optimal path, the cubic uniform B-spline curve is introduced to improve the smoothness of the path. Compared with the traditional algorithm, the path planning experiments in two different scale environments show that the proposed algorithm reduces the number of turns by 55. 6% and 87. 5% , respectively. The convergence speed is improved by 87. 5% and 100% , which verifies the superiority of the proposed algorithm. Finally, taking QBot2e as the platform, the algorithm is applied to indoor automated guided vehicle (AGV) path planning to further evaluate the practicability of the algorithm.