Session: CIE-14/15-01 CAPPD: Joint Topics

Paper Number: 143753

143753 - Multi-Task Learning for Intention and Trajectory Prediction in Human-Robot Collaborative Disassembly Tasks 

Human-robot collaboration (HRC) has become an integral element of many industries, including manufacturing. A fundamental requirement for safe HRC is to understand and predict human intentions and trajectories, especially when humans and robots operate in close proximity. However, predicting both human intention and trajectory components simultaneously remains a research gap. In this paper, we have developed a multi-task learning framework designed for Human-Robot Collaboration (HRC), which has two types of input: robot and human motion sequences. These inputs are processed through a bidirectional Long Short-Term Memory (Bi-LSTM) network as an encoder, which extracts a latent representation of the data. Then, the framework branches into two decoding tasks. The first task utilizes a Bi-LSTM combined with a Support Vector Machine (SVM) to identify the operator's intended path. On the other hand, the second task employs a Bi-LSTM followed by a dense neural network layer to forecast the movement trajectory. The proposed framework uses multi-task learning (MTL) to understand human behavior in complex manufacturing environments. Key innovations include the application of an attention mechanism to capture temporal dynamics in human motion sequences and a comparative analysis of various encoder architectures. We validate our framework through a case study focused on a human-robot collaborative disassembly task. The findings confirm the system's capability to accurately predict both human intentions and trajectories.

Presenting Author: Xinyao Zhang University of Florida

Presenting Author Biography: Cynthia (Xinyao) Zhang is a PhD student in Environmental Engineering Sciences at the University of Florida.

Authors:

Xinyao Zhang University of Florida
Sibo Tian Texas A&M University
Xiao Liang Texas A&M University
Minghui Zheng Texas A&M University
Sara Behdad University of Florida