In smart city, the deployment of network nodes of wireless networks has direct effect on network quality of service. This problem can be described as deploying appropriate AP as access nodes and special nodes as gateway nodes to aggregate traffic to Internet in a given geometric plane. In the paper, wireless mesh network as an example, number and deployment location of AP nodes can be determined by the regional flow of people statistics, and gateway nodes deployment is abstracted as a geometric K-center problem. To solve the geometric K-center problem, an improved adaptive PSO algorithm is proposed to optimize the minimum coverage radius. The fitness function is redesigned, and random inertia weight adjustment, adaptive learning factor, neighborhood searching strategy are introduced to the improved PSO to get wider solution. Compared with GA algorithm and K-means algorithm, simulation results show that the improved PSO algorithm is more stable and can get shorter path length, thus the network quality of service can be improved.