There are a large number of solder spots in the welding process of robot body in white. How to use intelligent algorithm toplan the path is an urgent problem to be solved. This article transforms the welding path into a TSP problem. Through genetic algo-rithm, the path planning of welding robot is realized. In this paper, we compare the evolution speed of genetic algorithm under differ-ent crossover length, and get the conclusion that the shorter the crossover length, the faster the evolution speed.