Design and Implementation of a Computer Vision System for Robotic Disassembly of Electric Vehicle Battery Pack

The increasing demands for greener transport solutions results in a foreseeable increase of Electric Vehicles (EV). The development of EVs puts strong demands on the development of LithiumIon Batteries but also into its dismantling process, which today is made manually and therefore is time-consuming. LIBRES, Lithium-Ion Battery Recycling, is a project owned by Norsk Hydro with the goal of finding fully-automated solutions for complete LIB batteries dismantling. The work presented in this thesis is a part of the LIBRES project and its aim is to develop a task planner for EV LIB battery pack dismantling to a module level through the design and implementation of a computer vision system. The designed task planner integrates the communication between the primary system elements (i.e. the robot, the 3D camera and the computer). The aim of the task planner is (1) to scan the dismantling scene, (2) to identify the different LIB components and their locations in the battery, (3) to create a dismantling order and lastly (4) to move the robot to the detected dismantling positions. The object detection algorithm "You Only Look Once" combined with the projects pose estimation part is used to find the different components in the disassembly scene and determine their position. To implement the designed task planner, a Volkswagen Hybrid LIB pack is considered. Results show that the method has a good accuracy with errors lower than 5 mm. In addition, the system is able to perform steps (1), (2), and (3) mentioned above, in an average time of 40 seconds.