Method for Bottle Opening with a Dual-Arm Robot.

Area of Science:

• Robotics
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Individuals with injuries or disabilities often require assistance with daily manipulation tasks.
• Robotic assistance can significantly enhance independence and quality of life.

Purpose of the Study:

• To develop and validate a robotic method for automated bottle opening using a dual-arm robot.
• To enhance accessibility for individuals needing help with common tasks.

Main Methods:

• Utilized the TIAGo++ dual-arm robot with an RGB-D camera for vision-based object detection.
• Implemented a control scheme combining position, force control, and Deep Reinforcement Learning (DRL) for manipulation.
• Planned safe trajectories for grasping and unscrewing actions.

Main Results:

• Successfully trained a Deep Reinforcement Learning agent for precise gripper rotation control.
• Demonstrated effective bottle opening through coordinated dual-arm manipulation and force feedback.
• Validated the system's practical viability in experimental settings.

Conclusions:

• The proposed robotic approach effectively automates bottle opening, enhancing accessibility.
• Integration of vision, manipulation, and DRL offers a robust solution for assistive robotics.
• The TIAGo++ robot shows promise for real-world applications in aiding individuals with limited mobility.