A modeling method for robots is proposed, in which a convex polyhedron is represented as a set of inequalities and a robot is represented as a union of convex polyhedrons. With this method, collision between robots can be detected by solving a set of linear programming problems at every sampling instant. By detecting possible collision at every sampling instant, a directed graph for robots is created. The motion planning problem of robots is then transformed into a path searching problem in the directed graph and can be solved by exisiting searching algorithms.