Robotic Mapping and Localisation in Feature Sparse Water Pipes

Water, a highly precious resource, is distributed to buildings by networks of pipes. As pipe materials age they are prone to damage, which can cause wastage of water and bacterial infiltration. Water distribution systems therefore require inspection, maintenance and repair. However, water pipes are usually buried and so are difficult to access. Robotic systems have great potential for inspecting these inaccessible pipelines but whilst there are many techniques for robot pipe inspection itself, an as yet unsolved problem is accurately locating damage in pipes once found. Accurate localisation of damage would greatly improve maintenance: speeding repair, reducing disruption and thereby reducing costs. The localisation problem can be addressed by robot navigation algorithms, specialised to water pipes, which is the focus of this investigation.

There are a number of challenges for robot navigation in water pipes. Firstly and most importantly the water pipe is a feature sparse environment. Most current robot navigation systems deal with indoor and outdoor environments, which contain numerous landmark features. However, pipe walls lack features that can be used as landmarks. Secondly, in pipes standard range and bearing sensors can only detect features that are nearby due to the close and enclosing proximity of the surrounding pipe wall. Thirdly, unlike indoor or outdoor navigation, the in-pipe robot has a very restricted route (moving either forward or backward), which limits the perspective of the robot on landmark features. Therefore, robot navigation in water distribution pipes is a difficult problem.

In this paper we develop a novel method for mapping and localisation in water pipes using voids outside the pipe wall as features, detected by ultrasonic scanning through the pipe wall. We demonstrate the approach experimentally in a laboratory environment. The experimental results show that ultrasonic sensing can be used to successfully build a map based on soil depth outside the pipe wall, and the navigation algorithm can be used to localise using voids as features.