抄録
This paper proposes a multi-robot system to estimate the position of an unknown moving object and track it. Concerning tracking mobile objects, it is important to know where the target is. To recognize the target, we consider attaching a monocular camera to each robot. The target’s position is calculated in two directions of the target from robots. However, since the camera’s information includes observation noises, it is difficult to use simply calculated positions to track the target. In this research, we employed an extended Kalman filter to estimate the position of a moving object. In addition, the estimation accuracy was improved by introducing pseudo-measurement. Each robot track target by optimal control based on the differential game theory. If the target and robots align, the noise is amplified because of triangulation. To avoid this situation, we employed virtual springs to control multi-robots formationFinally, numerical simulations in a two-dimensional plane were performed to verify the usefulness of this method.
