Multi-sensor Fusion Incorporating Adaptive Transformation for Reconfigurable Pavement Sweeping Robot

Abstract

An efficient sensors fusion framework in an autonomous robot is necessary for various functions like object detection and perception enhancement. Multi-sensor calibration techniques are used to fuse multiple static sensors into a single frame of reference. However, for reconfigurable robots, sensors can change pose during reconfiguration need a robust adaptive sensor fusion to account for the relative change in sensor position and orientation. We propose an adaptive sensor fusion framework that can be implemented on any reconfiguration robot to adjust calibration parameters. Our paper formulated an adaptive sensor fusion method, implemented it in real-time on an autonomous reconfigurable pavement sweeping robot called Panthera, and demonstrated qualitatively the accuracy of the proposed sensor fusion framework for environment perception during robot reconfiguring on the pavement.