Moving Target Tracking of Omnidirectional Robot with Stereo Cameras

Abstract

The omnidirectional mobile robots have attracted great attention in last twenty years. Their major advantage superior to the traditional car-like mobile robots, whose motion is subject to the nonholonomic constraints, lies on the feature that their linear and rotational motions can be simultaneously and independently carried out. A typical mechanical structure of omnidirectional robot utilizes three universal driving wheels which can be either driven or slid along any direction [2]. This special motion pattern significantly simplifies the path planning and motion control tasks. For the motion control of omnidirectional mobile robots, the majority of work so far, however, only consider the kinematic model for motion control. This is equivalent to assuming that the robots are massless bodies and therefore can ideally respond to the input motion control commands, which indeed does not reflect the real situation especially for heavy and fast moving robots. Due to this reason, efforts have been made to develop precise dynamic model to improve robots' performance [1]-[3]. In recent years, the visual servoing which combines control application with vision systems has become a hot topic. However, only a few studies [12], [13] combined the vision system with omnidirectional robot which has excellent maneuverability. In this study we integrate the motion control with stereo vision for target tracking. Unlike previous work [4], [9], in which control laws were based on image space using the inverse image Jacobian matrix for motion estimation, our algorithm is to directly estimate the relative position and velocity errors in Euclidean space. By using visual feedback, a simple PD controller is proposed. The PD control ensures the asymptotic stability of the tracking error if the target is still which is equivalent to the case of docking, or of the bounded-inputbounded-output (BIBO) stability if the target is moving with varying but bounded rotational and linear velocities. The paper is organized as follows: Section 2 presents the robot motion equation and tracking task. The dynamic model of the overall system is given in Section 3. The vision based relative motion and tracking error estimation is introduced in Section 4. The PD control law is proposed in Section 5. As a practical justification, the experimental tracking results obtained are given in Section 6. Finally, conclusions and future work are presented in Section 7. O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m