Abstract: When multi-robot exploration systems are teleoperated by human, they should follow the operator's intension; on the other hand, they should satisfy the network connectivity constraints and improve their own performance like communication efficiency. This paper focuses on coordinating the relationship between the above two aspects. By attenuating the frontier grids sampling rate and using the teleoperator interest oriented task allocation algorithm, the task allocation decisions and the operator's intention are coordinated by the real-time task allocator to solve the real-time requirements and human-robot decision coordination. In topology controller, an optimization Steiner tree problem is solved to minimize the number of Steiner points and maximize the weight sum of the other points, and thus the connectivity constraints are satisfied. In the selector for map fusion center, breadth first search is adopted to reduce transmission bandwidth cost and energy consumption. In typical indoor environments, comparative simulations against fully autonomous system ''Possible Moves Sampling'' and comparative experiment against an exploration system without a map fusion center are performed. The results show that the proposed system has higher and stabler exploration efficiency and lower data transfer amount, which proves the effectiveness of the human-robot shared control method.