Large language model-guided graph convolution network reasoning system for complex human-robot collaboration disassembly operations
Jinhua Xiao, Sergio Terzi
Abstract
This paper proposes a systematic reasoning framework combining Large Language Models (LLMs) with Graph Convolutional Networks (GCNs) for Human-Robot Collaboration (HRC) reasoning system for the actual disassembly operations. GCNs with the sensor data can be used to capture spatial-temporal relationships for human operation actions, while LLMs interpret text-prompt data inputs to provide the contextual insights that focus on relevant aspects of human-robot collaboration disassembly. The system enables the real-time reasoning of disassembly tasks and disassembly strategies, thereby enhancing both the safety and efficiency of HRC disassembly operations. This approach is particularly valuable in complex and dynamic environments, where the capability to quickly and accurately understand the complex disassembly tasks based on HRC disassembly environment that is crucial for successful collaboration disassembly between humans and robots. To address key challenges in ensuring the safe, and efficient disassembly operations, it is necessary to provide the way for more advanced human-robot collaboration disassembly with the complex disassembly environments in the disassembly operations.