Multisensor fusion in Reconfigurable Robots using Adaptive Transformation

Multisensor fusion is essential for various applications like accurate autonomous navigation, object detection, and perception. The calibration approach for getting accurate and meaningful results utilizes the intrinsic and extrinsic calibration of sensors, given that the sensors are fixed relative to each other. The robots with the ability to change their morphology are referred to as reconfigurable robots. The reconfigurable robot named Panthera is taken as a use case to showcase the need and practical application where the relative change in the position of the sensors occurs. The proposed approach account for the adaptive transformation, which captures the change in the relative pose of the sensors in real-time. The scheme presented is demonstrated using multiple stereo cameras and three-dimensional LiDARs. Moreover, the result for the effectiveness of the adaptive transformation for two cases is presented and discussed. The framework is generic and suitable for a multisensor fusion where each sensor is subjected to change in relative pose.