In the current method of optimizing the design of spatial building structures, the high convergence of the algorithm makes it impossible to realize multi-objective population optimization, and it is easy to fall into the local optimal solution. The method has the problems of low optimization quality, high optimization cost, and low seismic performance. To solve the above problems, in this work, a method of optimizing the design of spatial building structures based on the improved particle swarm algorithm is proposed, where the seismic performance and engineering cost of space structure are the optimization target. A co-evolutionary multiple sub-groups mechanism was introduced to solve the problem of population optimization among multiple objectives in the optimization of the seismic design of spatial structures. Meanwhile, the elite learning strategy was introduced to improve the particle swarm algorithm. The optimal design scheme satisfying the objective function was selected, and the optimization of the building space structure with seismic constraints was completed. The experimental results showed that the proposed method is characterized by high optimization quality, low optimization cost, and high seismic performance.