PIRATE is a multi-link, snake-like robot designed at the RAM group at the University of Twente for the aim of inspecting industrial pipes in the Netherlands. Currently, the project is being worked towards the implementation of an autonomous version of the robot.
One of the existing problems is how to robustly navigate the robot to go through sharp corners in the pipe due to the fact that those corners do not have a certain turning angle and the diameters of the pipes can also be different so it is hard to develop a hand-engineered approach which can tackle all of the situations.
One possible solution is to employ a learning-based motion planner for the robot by applying Reinforcement Learning (RL). RL is a machine learning technique that tries to find an optimal policy for an agent to achieve certain tasks by continuously interacting with the environment and improving the policy with gained experience.
Therefore, this project focuses on designing a robust, high-level path-planner for PIRATE by leveraging Reinforcement Learning in order to solve the navigation problem of PIRATE at sharp pipe corners.