Non-Oscillatory Multi-Robot Motion Planning for Stable Target Capture

Abstract

This dissertation presents a motion planning method to capture a stationary target without oscillations. Potential field method, which is applied to the proposed method, is a popular robot motion planning method because it is simple and easy to implement. Oscillation, which is an inherent problem in the potential field method, can cause unnecessary motions and system instability. Capturing motion can be divided into two parts: the motion of approaching the target and the motion of enclosing it. In the enclosing motion, the repulsive forces between the robots are the main causes of the oscillation, and the robots should surround the target at regular intervals. Therefore, the approaching and enclosing motions are of different character. Hence, we have separated the capturing motion into two phases (approaching and enclosing phases) and have proposed an oscillation prevention method for each phase. The oscillation in the approaching phase can be removed by controlling the direction of the total force acting on the robot. As the total force is the sum of the repulsive and attractive forces, the direction of the total force can be changed by controlling the magnitudes of the forces. This dissertation proposes a method of obtaining the coefficients of the repulsive and attractive forces to prevent oscillations in the vicinity of obstacles and in the presence of other moving robots. If a robot enters the region within a certain range from the target, its phase changes into the enclosing phase. In this phase, the polar coordinate is introduced so that the robots would be at regular intervals. A potential trench is formed on the r-axis so that the robots converge on an enclosing circle. On the θ-axis, a goal angle is assigned to each robot so that the intervals between the robots are regular. We have proved that the robots converge on their goal points without an oscillation. As the previous studies on the oscillation problem were for a single-robot system, this is the first study on the oscillation problem for a multi-robot system. In the simulations, the proposed method showed better performance in terms of the number of oscillations and arrival time compared with the conventional potential field method, which uses arbitrary coefficients.