The velocity obstacle (VO) method is one of local path generation method considering a velocity of obstacles. By dividing an available velocity region into collision and collision-free area, a robot can avoid collisions using the VO. However, if there are numerous obstacles near a robot, the robot will have very few velocity candidates. In this paper, a method to choose an optimal velocity by introducing a cost function about safety of the velocity, and the cost function consists of a pass-time and a clearance. By latticizing available velocity map of a robot, each velocity can be evaluated from the cost function and a robot can select better velocity among collision-free velocity candidates. A performance of introduced method is compared to other VO method using simulation, and experiments are conducted to verify the results of simulation.