University of Twente Student Theses

Login

Constraint Based Trajectory Planning of a Robotic Manipulator in Nuclear Power Plant Using Deep Reinforcement Learning

Mughal, Maham Ehsan (2025) Constraint Based Trajectory Planning of a Robotic Manipulator in Nuclear Power Plant Using Deep Reinforcement Learning.

[img] PDF
40MB
Abstract:Nuclear power is essential for meeting the climate change goals as it produces carbon free electricity. But the highly radioactive environment inside a nuclear power plant makes it very dangerous for humans to maintain it. An important component of a nuclear power plant that needs frequent maintenance is a steam generator that con- verts water into steam and acts as a heat exchanger. In this thesis, a manipulator named UR16e from Universal Robots is used to reach the position where mainte- nance is required inside the steam generator compartment using deep reinforcement learning. The steam generator compartment has a narrow opening called a manway to enter it which makes the environment confined. A reward function is designed to achieve the desired position and orientation of the end-effector without collisions between the robot and the steam generator walls using proximal policy optimization algorithm. With simulation results, it is shown that a multi-objective reward func- tion leads to unintended results. Using hierarchical reinforcement learning in this problem improves the results by breaking down the original goal into subgoals and designing separate reward functions for each subgoal. Lastly, curriculum learning is used to achieve faster convergence by exposing the agent to increasingly complex environment using dynamic reward functions.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general
Programme:Robotics MSc (60973)
Link to this item:https://purl.utwente.nl/essays/106237
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page