Trust-Based Optimal Subtask Allocation and Model Predictive Control for Human-Robot Collaborative Assembly in Manufacturing

We develop a one human-one robot hybrid cell for collaborative assembly in manufacturing. The selected task is to assemble a few LEGO parts into a final assembled product following specified instructions and sequence in collaboration between the human and the robot. We develop a two-level feedforward optimization strategy that determines the optimal subtask allocation between the human and the robot for the selected assembly before the assembly starts. We derive dynamics models for human’s trust in the robot and the robot’s trust in the human for the assembly and estimate the trusts. The aim is to maintain satisfactory trust levels between the human and the robot through the application of the optimal subtask allocation. Again, subtask re-allocation is proposed to regain trusts if the trusts reduce to below the specified levels. Furthermore, it is hypothesized that fluctuations in human’s trust in the robot may cause fluctuations in human’s speeds and the human may appreciate if the robot adjusts its speeds with changes in human speeds. Hence, trust-based Model Predictive Control (MPC) is proposed to minimize the variations between human and robot speeds and to maximize the trusts. Experiment results prove the effectiveness of the hybrid cell, the feedforward optimal subtask allocation and of the trust-based MPC. The results also show that the overall assembly performance can be enhanced and the performance status can be monitored through a single dynamic parameter, i.e. the trust.