Gaze Control for Active Visual SLAM via Panoramic Cost Map

In this work, we aim to improve the positioning accuracy of the visual simultaneous localization and mapping (VSLAM) through actively controlling the gaze of the positioning camera mounted on an autonomous guided vehicle (AGV). A panoramic cost map (PCM)-based gaze control method (PGC) is proposed for the active VSLAM. Different from traditional method, a panoramic camera is added beside the positioning camera to aid the gaze control of the positioning camera. The panoramic camera is used to perceive the environment and evaluate the potential performance of each available orientation of the positioning camera. The evaluation of all the available orientations will make up a panoramic cost map. The cost map is then used to help the gaze control method to select an optimal target gaze for the positioning camera. In the calculation of the panoramic cost map, the effective factors of the VSLAM, such as feature points and moving objects, are taken into consideration. In the gaze control method, we also take into consideration of errors of the system, the time delay of the proposed method, and the velocity of the AGV. The test results in different scenes with different VSLAM algorithms show that the proposed method can improve the positioning accuracy of all the tested VSLAM algorithms compared to fixed camera gaze.