The formation of intelligent platforms is a multi-objective constraint problem. It is necessary for the multi-agent to automatically generate the path from the initial position to the specified end point, and at the same time meet the constraints of space collision avoidance on the multi-target intersection path and the constraint of the agent’s motion ability, and pursue the shortest formation time of the whole formation as much as possible. In this paper, we propose an improved method for obtaining the starting set of multi-agent formation based on particle swarm optimization. First, the starting point set of the formation is defined, and then the particle swarm optimization algorithm is selected as the optimization algorithm to find the best starting point of the formation. Then the turning performance of the intelligent platform is studied, and the assembly route of each agent is designed by combining the Dubins curve. The simulation results show that the proposed method avoids collision between agents, shortens the time of agent formation assembly, and effectively ensures the reliability of the assembly route.