According to the traditional path planning algorithm in robot path planning without considering the mechanical arm interference, low searching efficiency and unreasonable path, an improved genetic algorithm is proposed based on the establishment of manipulator space model and the analysis and calculation of manipulator operation space by using two projection method after dividing barrier model and operation space into cubic mesh. In the design of genetic a lgorithm, methods of using combination of real number coding and 3D coordinate coding,improving crossover operator and defining the minimum gene crossover ratio improve the search efficiency and simplify the calculation process. The simulation model of manipulator is established by Matlab, and the validity of the algorithm is verified. Compared with the traditional path planning algorithm, the search efficiency and the optimal path ratio have been significantly improved under the same experimental conditions and experimental objects.