In order to improve the response ability and adaptability of robot system to complex instructions, a human-multi-robot shared control based on surface electromyography (sEMG) and eye tracking technology is proposed. The self-designed EMG wrist strap is used to collect the surface EMG signals of the operator's forearm in real time, and the lightweight convolutional neural network is designed to recognize typical gestures and map them into system control instructions. The eye tracker is used to obtain the operator's gaze point information and gaze trajectory, and the robot is selected and the desired path is drawn. Two new interaction modes are integrated to construct a human-computer interaction platform, and a human-multi-robot shared control system is constructed with multiple small robots. Experiments are carried out on a human-multi-robot hardware-in-the-loop simulation platform, and the results show that the shared control method is feasible and effective.