Abstract: According to the social development requirements and the state-of-the-art technology, an M2M2A (man-or machine-to-machine to actuation) solution is proposed for multi-robot collaboration, which can implement environmental adaptability, man-machine integration and intelligent decision in industrial applications. Firstly, the structure and organization of multi-robot operating system consisting of humans, intelligent center, robots and sensors are designed. A hybrid M2M communication model of C/S (client/server) and P2P (peer-to-peer) is constructed, in which the C/S model can realize remote transmission of control commands and machine information between man and machines, and the P2P model can accomplish information sharing among those machines. A series of intelligent modules for robot self-localization, environment perceiving, collision-free path optimization, trajectory planning and motion control are developed to perform autonomous operation of robots. An open and modular M2M2A system is developed, in which the communication module, intelligent decision module and control module are integrated into collectives by datastream. A typical experimental system for multi-robot collaborative assembly is designed. By the communication module, the environmental information, the robot pose or position information can be transported or shared in real time. By the robot intelligent modules, the robots will reschedule their working time sequences, and then replan their paths or trajectories to optimally accomplish the task once the surrounding changes.