For the structure characters of the products, a matrix of geometric constraints is established. An optimization model of assembly sequence planning is studied to maximize the number of assembly operations with the geometric assembly constraints and minimize the number of changing assembly direction. The updating rules of particle position and velocity are derived based on the value transformation. A heuristic algorithm with randomness characteristic is provided in order to generate the initial populations. A particle swarm optimization algorithm with the depth local search strategy is developed to solve the assembly sequence planning problem. To evaluate the performance of the algorithm and the quality of the assembly sequences, an experiment of an assembly instance is tested. The results show the effectiveness and stability of the proposed algorithm on solving the assembly sequence planning problem.