In view of high dimension, the difficulty of training, the problem of slow learning speed in the application of BP neural network in mobile robot path planning, an algorithm of reinforcement Q learning based on online sequential extreme learning machine (Q-OSELM algorithm) was proposed in this paper. And then, due to the random selection of weight and threshold parameters, it also proposes an extreme learning machine algorithm optimized by particle swarm (PSO-ELM algorithm) in this paper. Firstly, Q-OSELM algorithm obtains current environment and the status information of the robot through the characteristic of reinforcement learning, which combines dynamic network with supervised learning. After that, the online sequential extreme learning machine is used to approximate the function of the current status to get the rewards and punishments of the current status; Secondly, it is used to solve the problem of slow training speed by the characteristic of less parameter settings and better generalization performance. PSO-ELM algorithm is used to optimize the input weights and the hidden layer bias of the extreme learning machine which have been seen as the particle of particle swarm optimization algorithm to improve the network structure of the extreme learning machine. It could overcome inaccuracy of traditional extreme learning machine through particle swarm optimization algorithm. Finally, the performance of two learning algorithms is verified. The simulation experimental results show that the Q-OSELM learning algorithm improves the initiative of machine learning. And compared with the Q-OSELM algorithm, the PSO-ELM algorithm has better generalization ability and higher training precision. Simulation experiments are carried out to verify the stability and convergence of the two algorithms.