IMU and cable encoder data fusion for in-pipe mobile robot localization

Abstract

Inner pipe inspection of sewer networks is a hard and tedious task, due to the nature of the environment, which is narrow, dark, wet and dirty. So, mobile robots can play an important role to solve condition assessment of such huge civil infrastructures, resulting in a clear benefit for citizens. One of the fundamental tasks that a mobile robot should solve is localization, but in such environments GPS signal is completely denied, so alternative methods have to be developed. Visual odometry and visual SLAM are promising techniques to be applied in such environments, but they require a populated set of visual feature tracks, which is a requirement that can not be fulfilled in such environments in a continuous way. With the aim of designing robust and reliable robot systems, this paper proposes and evaluates a complementary approach to localize a mobile robot, which is based on sensor data fusion of an inertial measurement unit and of a cable encoder, which measures the length of an unfolded cable, from the starting point of operations up to the tethered robot. Data fusion is based on optimization of a set of windowed states given the sensor measurements in that window. The paper details theoretical basis, practical implementation issues and results obtained in testing pipe scenarios.