To improve the localization accuracy of angle of arrival in passive troposcatter surveillance system, an improved particle swarm optimization was employed to arrange distributed nodes. Geometry dilution of precision was deduced and treated as the evaluation index. Chaos theory was adopted to modulate the position parameters for particle swarm optimization. Adaptive inertia weigh and learning factors were introduced to improve the optimization ability. In order to avoid getting stuck at local optimization and enhance the global exploration ability, swarm optimization based on two subgroups was adopted. The crossover operation and mutation operation were carried out based on two subgroups as well. Simulation results demonstrate that the proposed algorithm can obviously improve the location performance of passive troposcatter surveillance system, and the operating time is less than that of the traversal optimization.