Abstract: The local path planning of mine rescue robot based on particle swarm radial basis function network is proposed. The rescue robot is trained by using the algorithm to simulate the complex environment of mine; adjusting weight, the optimal solution is obtained. At the same time, the deterministic local planning algorithm is used to optimize the particle swarm algorithm to make it more reasonable for local processing. Studies show that the algorithm avoid the least-square method easily trapped in local minimum problem, and can use a very fast speed to approximate the optimal solution, the optimization of the result is more reasonable, more practical and more accurate.