Abstract: The hunting task for multiple mobile robots in an unknown environment is modeled as five states, namely, lining-up, random-search, surrounding, catch, and prediction states in the paper. The state transition conditions, and the line-up, search, outflank, catch, predict and direction-optimization strategies are proposed to ensure smooth fulfillment of the task. The target to be captured, called the invader, is endowed a kind of safety motion strategy thus the hunting is more difficult. We illustrate the feasibility of the approach by the simulation.