SPREKER
Edwin Dertien, promovendus, UT
TITEL
World modeling for a pipe inspection robot
ABSTRACT
In the Netherlands, the urban low pressure gas distribution mains stretches 100.000 km. At present above ground ‘sniffing’ methods can locate leaks, but no quality information of the underlying net is gathered. Hence the project goal to build an autonomous robot for in-pipe survey, capable of autonomous inspection and detection of leaks, cracks and deformation.
A robot has been developed, capable of moving through the pipes found in the urban low pressure distribution net, as well as most of the types of obstacles that can be found (bends, T-joints, etc). This robot consists of two driven modules and a number of payload modules. In order to make (autonomous) navigation possible, the control system of this robot has to derive a model of the world it is driving trough, based on a-priori data (network map) but mostly on sensor information.
The information available for modeling the robot’s surroundings are the robot’s internal state data, inertial measurement data, odometric data and information from a vision system. This vision system consist of a solid state conical laser projection combined with a standard resolution camera. By comparing this vision data to simulation-generated ‘template data’ the system is able to distinguish and characterize the various types of obstacles inside the pipe network.
All sensory information of the robot is collected in a simulation environment on a separate computer system, where the world model is generated. Based on this worldmodel the robot will eventually be able to navigate through the network, avoid obstacles, etc. The next step is to control the robot from the same simulation environment, so an effective ‘hardware in the loop’ simulation takes place.
This Pirate project (pipe inspection robot for autonomous navigation) is developed by a consortium consisting of the Control Engineering group at University of Twente, Kiwa Gastec and a number of gas distribution net operators.