BRITE Pivot: Deep Robotic EV Battery Repair: An LLM-Powered Task-Motion-Manipulation Planning Framework

This BRITE Pivot award supports research that contributes new knowledge and novel robotic solutions related to the repair of aged electric vehicle (EV) batteries, thereby promoting the progress of science, and advancing prosperity and welfare. EV battery repair currently relies heavily on skilled manual labor, requiring extensive expertise and effort just to dismantle a battery before the repair process can even begin. This challenge arises from the complex design and the uncertain conditions of aged batteries, as well as the significant safety risks associated with handling high-voltage components. This project seeks to solve this challenge by leveraging emerging artificial intelligence (AI) technologies, especially large language models (LLMs), to significantly enhance robotic capabilities. By providing an efficient robotic repair solution that is both economically viable and safe for human workers, this project has potential to extend the automotive lifespan of EV batteries, manage the growing volume of aged EV batteries, address skilled labor shortages, maximize the use of critical materials before recycling, and promote a sustainable, long-term, domestic, and circular battery supply chain. A series of workshops and webinars will be organized to provide training opportunities for next generation workforce in robotic EV battery repair and remanufacturing. Existing automation solutions for EV battery repair are highly customized, expensive, and often require extensive reconfiguration and human intervention to address the associated complexities and uncertainties. This research aims to create novel battery repair solutions with greater flexibility and adaptability in robotic task planning, motion planning, and manipulation by leveraging emerging AI technologies. The project looks to develop a fine-tuned, multimodal LLM tailored for robotic EV battery repair. This model seeks to enable more flexible, human-like task planning beyond rigid, pre-programmed sequences that struggle to handle complex battery repair processes. Furthermore, the research seeks to create a new motion planning framework to incorporate real-time human guidance for coordinated planning among heterogeneous robots. The research also includes activities that look to design a specialized robotic gripper, integrated with a modified residual reinforcement learning algorithm, to handle interlocking structures commonly found in EV batteries. Collectively, these foundational advancements seek to enable adaptable EV battery repair encompassing disassembly, replacement, and reassembly, and will lay the groundwork for a new paradigm in the EV battery remanufacturing industry. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.