The work aims to optimize the AGV routing problem of logistics. An improved chaos particle swarm optimization algorithm was proposed. The route planning model based on Bézier curve was adopted to significantly improve the length and smoothness of the AGV trajectory by adjusting the number of control points of the Bézier curve. Chaos particle swarm optimization (CPSO) algorithm was applied to optimize the control points of Bézier curves. The fitness function was introduced to assess whether the termination criteria were met. If the maximum iteration number was achieved or the optimal solution was not modified when the number of iterations was given, the CPSO algorithm was terminated. Finally, the selected control points were used to calculate the shortest and smoothest trajectory, which improved the optimization capacity of the algorithm. Bézier curve initialized by CPSO algorithm can obtain the smoother shortest path.