Interference between physical objects in the workspace and the moving human body may cause serious problems, including errors in manual operation, physical damage and trauma from the collision, and increased biomechanical stresses due to movement reorganization for avoiding the obstacles. Therefore, a computer algorithm to detect possible collisions and simulate human motions to avoid obstacles will be an important tool for computer-aided ergonomics and optimization of system design in the early stage of a design process. In the present study, we present a method of modifying motions for obstacle avoidance when the object intrudes near the center of the planned motion. We take the motion modification approach, as we believe that for a certain class of obstacle avoidance problems, a person would modify a pre-planned motion that would result in a collision to a new one that is collision-free, as opposed to organizing a totally unique motion pattern. Our method assumes that there is an existing motion sample that satisfies the initial and final hand position constraints. This existing motion sample is called a root motion. If the root motion is found to incur a collision with obstacles, it will be modified to become collision-free. The modification is not done at random, but guided utilizing a “via” posture (Rosenbaum et al., 1999). This idea is based upon the hypothesis that the human brain can reorganize a pre-planned motion to avoid obstacles by locating a “via” posture at a certain time of the motion and altering the pre-planned angular motion trajectories to pass through the “via” posture. An optimization model is presented for finding a via posture via which the resulting modified motion becomes a realistic variation of the root motion as well as collision-free. Our initial simulation results indicate that the proposed method is promising for planning and simulating obstacle avoidance motions.