Humans working alongside robots is becoming more prevalent, but training human employees to work with robots can be difficult and time consuming. Researchers from the University of Georgia have created a new virtual reality space to help make training faster and easier.
One task that often requires robotic help is disassembly. It’s critical for recycling parts and valuable materials from electronics nearing the end of their lifespan. However, disassembly tasks often present their own challenges. “Unlike assembly, which has a very standard procedure, disassembly is slightly more complicated,” said Beiwen Li, corresponding author of the study and an associate professor in UGA’s College of Engineering. “It may not work out the best if we just inverse the whole assembly procedure.”
The researchers developed VR Co-Lab to help ease confusion. By training employees digitally, not only can they practice disassembling recyclables without damaging materials, but they can also learn to avoid injury and collisions with the robot.
In the virtual space, users practiced taking apart a hard disk with a robot’s help. When users put on their headset, they found themselves in a virtual workstation like what they’d encounter on the job. They saw the various tools and machinery needed for disassembly, as well as the robot helping them.
During training, workers followed a step-by-step procedure for disassembling the hard disk. Human users handled more precise tasks like unscrewing or picking and placing small bolts, while the robot arm managed larger bolts and loose items. The program also provided feedback to the user, measuring how long it took to complete the session and how many mistakes users made.
“There are a lot of tasks. It requires a complicated training for workers, typically,” Li said. “So, if we have a VR system, that will be very helpful in shortening training time. It is much easier than having pages and pages of written documents to be read by the user.”
One key part of the VR system is body tracking. VR Co-Lab uses Meta Quest Pro, which uses its cameras to track upper body movements in the wrists, elbows, shoulders and torso. This helps the program plan the robot’s movements based on the user’s actions, preventing collisions and improving interactivity between the human and robot.
The program warns of potential hazards that could lead to injury, such as collisions with the robot arm, while teaching users how to avoid them. The system can also be used to figure out how quickly the robot arm can work without overwhelming the employee.
Li and his team are planning more extensive user testing in the future to ensure the system is helpful for a variety of skill levels and applicable to tasks outside of disassembling hard disks. Improving the training process will be valuable as using robots becomes more common.