University of Twente Student Theses

This website will be unavailable due to maintenance December 1st between 8:00 and 12:00 CET.

Design of a high-level control layer and wheel contact estimation and compensation for the pipe inspection robot PIRATE

Geerlings, N.M. (2018) Design of a high-level control layer and wheel contact estimation and compensation for the pipe inspection robot PIRATE.

[img] PDF
Abstract:Automation of pipe inspections by using robots has many advantages over manual pipe inspection, such as faster, more consistent and more accurate inspection. For this reason the pipe inspection robot PIRATE is being developped by the University of Twente. In this project the first goal is to design a high-level control layer, which should allow the robot to autonomously move through various types of pipe segments, without falling. The second goal is to estimate the wheel slip ratio and friction coefficient. The Magic formula and Gaussian processes are used to estimate their relation. The third goal is to design a traction controller, to ensure reliable odometry and prevent falls in a vertical pipe. The high-level control layer is designed based on the motion primitives. In order to structure the behaviour of the robot, five finite state machines are used. The joints and wheels can be controlled by position control or open-loop voltage control. A set of fourteen commands is used to control the robot. The high-level controller is implemented in ROS using C++, in the RobMoSys style. At the Skill level, nodes for motion primitives and for sequences of motion primitives are implemented. An extra detector node can trigger a sequence by using information from e.g. a camera. The traction controller incorporates velocity control and gravity compensation for the wheels, and a controls the clamping force for the bending joints. This controller is implemented in a Simulink simulation model of the PIRATE. To estimate the wheel slip and friction force, the PIRATE drives forward and backward in a 2D straight ’pipe’ and in a real straight pipe. The linear velocity is determined by a camera with a marker detection algorithm. To evaluate the autonomy of the robot, the robot has to drive autonomously through a 2D 90� mitre bend. In simulation the robot has to drive up and down in a horizontal and vertical pipe while using traction control. When driving, wheel slip ratio’s up to 0.3 are observed. Differences in slip ratio are observed when the movement of the robot is obstructed. For the friction coefficient large differences between the wheels are observed. The robot is able to move through the 2D mitre bend, with the setup tilted up to an angle of 28�, when commands are provided by the operator, as well as when the robot is triggered by the perceived location of the marker. Simulation shows that the smallest odometry error is achieved with the traction controller that has a velocity controller, but no gravity compensation, and either a controlled clamping torque or fixed joint torque. The relation between the wheel slip and friction coefficient can not be identified with either the Magic formula or the Gaussian processes, due to inconsistent behaviour of the robot. The software architecture successfully allows for driving through a mitre bend autonomously. The traction controller is able to select a proper clamping torque such that the PIRATE does not slide out of a vertical pipe in simulation.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Mechanical Engineering MSc (60439)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page