Vision and laser sensor data fusion technique for target approaching by outdoor mobile robot

The authors have been developing an outdoor mobile robot intended to provide increased traveling distance by autonomously negotiating and crossing a road crossing intersection while traveling along pedestrian sidewalks in an urban environment. In this paper, high precision navigation towards a pedestrian push-button box by a mobile robot for the autonomous activation of the button is presented. We show a dual sensor fusion technique using a monocular camera and laser range sensor with which an outdoor mobile robot can detect, localize and then accurately navigate towards a button box so that it could autonomously press the pedestrian push button in order to trigger the crossing sequence. The method involves determining the image formation of the target on the image sensor of the camera, using it to estimate the object position in the real world and then using data from the laser range sensor to acquire a precise location of the object relative to the robot and finally perform the path planning. A two-tiered validation system, one at the vision level and the second at the laser scan data level, detects inaccurate detections and results in a robust system. The proposed method is also applicable for any form of target approaching. Experimental results verify the efficacy of the system and concluding remarks are also given.