Researchers use table tennis to understand human-robot dynamics in flexible environments

A team of researchers, led by Matthew Gombolay, an assistant professor in the School of Interactive Computing and director of the Cognitive Optimization and Relational Robotics (CORE) Lab at Georgia Tech, is using the subject. ping pong sport to demonstrate that humans can’t always trust a robot’s explanation of its intended action.

They developed what is called a “cobot”, using ping pong to demonstrate potential areas a robot can work closely with mankind partners to complete the mission.

The robot, a Barrett WAM arm equipped with a camera and paddle, was trained through a machine learning process called imitation learning. Researchers have developed a system to give robots positive reinforcement for successful volleys and negative reinforcement for unsuccessful volleys.

“We have also trained our robot to be a safe table tennis partner,” says Gombolay. “We’ve taken advantage of previous work on ping-pong and ‘learn from demonstration techniques’, where a human can teach a robot a skill like how to hit a ping pong ball or simply let the robot play. the person performing the tasks for the robot.”

According to Gombolay, the project demonstrates the potential for robots to work closely with humans in terms of physical and social capabilities, an important step forward for collaborative robots. The development of intelligent systems that can collaborate with humans has a wide range of applications, from manufacturing to healthcare to education.

In some cases, the researchers found that human participants lacked trust in the explanations given by the robot and were therefore less likely to cooperate with it. One potential reason is a lack of trust that the robot may not have the same goals or motivations as its human counterpart. Study participants were more likely to trust the robot’s explanation when they felt that the robot had the same goals and motivations as them.

“If we can figure out how to allow humans and robots to work together safely in extreme cases, it will help us better understand how to support interactions in multiple environments.” different environments,” said Gombolay, a co-researcher emphasizing the importance of developing a robot. can communicate effectively with people in a way that builds trust and confidence. This is especially important in contexts where the consequences of a mistake or miscommunication can be severe, such as in emergency or healthcare situations.

“While choosing to work with a robot is ultimately an objective behavior and can change depending on the context or risk level of the interaction, ultimately this element of trust is important,” says Gombolay. is the main driver behind your behavior and decisions”. . “In practice, we often find that people design and implement impressive robotic solutions, but that robot is not designed to generate the right level of trust from the end user.”

The researchers suggest that one possible solution could be to design robots that are more transparent in their decision-making. By providing human partners with a clear understanding of how the robot arrives at a particular decision or action, it is possible to build trust and confidence in the robot’s abilities.

“The larger goal of the project is to understand how to design robots for close, fast-paced interactions in manufacturing, logistics warehouses, restaurant kitchens, and even indoors. Of course, we need to know. how to design physically secure systems, but we also need to know what users find intuitive and trustworthy—what makes users feel safe,” says Gombolay.

“Otherwise, these robots would never have come out of the lab to co-exist with humans. I believe our research answers key questions in helping to design robots for interoperability. interact with humans, particularly with regard to how robots communicate their intentions to their human counterparts. But, of course, the research opens up more exciting opportunities than previously thought.”

An approach that could eventually be a wave of productive human-robot collaboration in a variety of environments.

The article will be presented at The 18th Annual ACM/IEEE International Conference on Robotic Interaction with Robots (HRI), March 13–16, 2023, in Stockholm, Sweden.