A Novel Training and Collaboration Integrated Framework for Human-Agent Teleoperation.

Area of Science:

• Robotics
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Human operators face increasing physical and mental workloads in teleoperation.
• Subjective factors can negatively impact operator performance and lead to errors.
• Agent-based methods show promise but require human intelligence integration.

Purpose of the Study:

• To propose an integrated framework for human-agent teleoperation using reinforcement learning.
• To enable co-optimization of human and agent performance through bilateral training.
• To provide quantifiable training feedback and agent-based arbitration.

Main Methods:

• Development of a truncated quantile critics reinforcement learning-based integrated framework.
• Implementation of an expert training agent and a bilateral training process.
• Experimental comparison of human-human and human-agent cooperation modes.

Main Results:

• Subjects trained with the agent completed tasks faster and with higher success rates.
• Human-agent cooperation resulted in significantly less workload compared to human-human cooperation.
• The framework demonstrated efficient and quantifiable training feedback.

Conclusions:

• The proposed human-agent teleoperation framework effectively improves operator performance.
• Reinforcement learning facilitates co-optimization of human and agent capabilities.
• Agent assistance in teleoperation reduces task completion time and operator workload.