This paper proposes a collision avoidance algorithm for two mobile robots with independent goals in a same workspace. Using skeleton map, an environment is presented as a graph consisting of nodes and arcs. Robots generate the shortest set of paths using the combination of objective functions of the two robots. Path for collision avoidance of a robot can be selected among three class; the shortest path, detour, paths with a waiting time at safety points around crossing points. Simulation results are presented to verify the efficiency of the proposed algorithms.