As the existing multi-robot collision avoidance algorithm is of high computational complexity and poor applicability, a new algorithm based on the generalized potential field and the traditional artificial potential field-based algorithm, which takes into consideration the consult and desire of the robot, is proposed for the motion planning of multi-robot system. In this algorithm, a repellence potential field is built based on the obstacles in the physical space, a gravitation potential field is built based on the desire, and a generalized artificial potential field is built by the topological product of the repellence and the gravitation potential fields. By solving the motion trend of the mini- cells in the established artificial potential field, a road planning can be successfully implemented. Experimental re- sults indicate that, as compared with the traditional artificial potential field-based algorithm, the proposed algorithm is also simple in structure and is superior in terms of optimal path planning and mobile collision avoidance.