Abstract: In the process of building cluster generalisation understanding building displacement is a very important task.Based on the analysis of existing algorithms，we discuss a new adaptive real-coded genetic algorithm for building cluster displacement. In this algorithm the spatial relationships between buildings and streets or roads and the Gestalt principles in the graphical representation of buildings arrangement are taken into account，and in the design of fitness function multiple factors are also considered，putting the displacement results more in line with the cartographic norms and map perception rules. Experimental results show that the algorithm can accurately converge to the optimal solution when the displacement space is sufficient，and the displacement effect is satisfactory.