To solve the problems that the Shuffled Frog Leaping Algorithm (SFLA) path planning is easy to fall into local optima and optimization effect is poor, an improved SFLA was proposed. Euclidean distance and the best frog of the population were added into the original algorithms update strategy, and a method to generate a new individual with a adjustable control parameter instead of the original random update operation was proposed. This paper firstly transformed the problem of robot path planning into a minimization one, and then defined the fitness of a frog based on the position of target and obstacles in the environment. The robot selected and reached the position of the globally best frog in each iteration successively to realize the optimal path planning. In the mobile robots simulation experiment, compared with other algorithms, the successes number of the improved algorithm was increased from 82 to 98, and the path planning time was decreased from 9.7s to 5.3s. The experimental results show that the improved algorithm has stronger security and better optimization performance.