Abstract: A general scheme of cooperative hunting for a moving target by multiple mobile robots in continuous unknown environments is presented. Hunting consists of encircling the target and closing to it, and the encircling behavior is realized with reinforcement learning algorithm. States are clustered in order to reduce the state space, Q learning algorithm is used to get the table of Q values, then the available action is selected according to the Q value table. Hunting of mobile target is realized with synthesized behavior, obtained by summarizing the outputs of all behaviors weighted. Hunting effects in different conditions are verified by simulation, and the results show that environments, velocity relationships between hunter and prey, and the escaping strategies of prey all have their effects on the result.