A multi-robot complete coverage path planning algorithm based on rolling optimization and decentralized predator-prey models is proposed to solve the problem of poor performance of multi-robot performing coverage tasks. The raster map is used to represent the environment space of the job. And based on the raster map, the reward function of avoiding predators in the predator-prey algorithm is modified, the moving cost reward function and dead-zone backtracking mechanism are added to build a decentralized predator-prey model. The rolling optimization method is introduced to avoid the robot falling into local optimum. The cumulative reward value of the robot covered grid during the prediction period is used as the fitness function, and the optimal movement sequence is solved using the whale optimization algorithm (WOA). Finally, simulation experiments are carried out in different environments. Compared with the biologically inspired neural network algorithm (BINN) and the boustrophedon-A* algorithm (BA*), the average path length planned by the proposed method is reduced by 16.69%sim17.33% and 10.32%sim20.03%, respectively, which verifies the feasibility and effectiveness of the proposed method in multi-robot full coverage path planning.