Video robots are well known to help (or obstruction) people, but in the real world, they are beautiful less social skills. MIT’s CSAIL computer science researchers are trying to improve this by teaching robots how to integrate with other robots to achieve their goals, according to new paper. This research can improve the connection between robots in residential areas, for example, and help psychiatrists to better measure interpersonal interactions.
To study this interaction, the researchers created a 2D simulation that allowed real robots to follow group and physical goals. For example, the physical goal may be to go to a tree somewhere on the grid, while the human goal is simply to focus on what another robot is trying to do and do the same, “such as helping another robot irrigate a tree,” according to CSAIL.
The robot is rewarded for actions that are close to its goals, with the same reward for support and a different reward for the obstacle. The team developed three types of robots: The first has only physical goals, the second has physical and team goals, but assumes that all robots have only physical goals. The third assumes that all others have social and physical goals, so they can perform high-level tasks such as socializing with others to achieve the goal.
Although young children seem to understand social interactions such as helping and hindering, we do not have a machine that can think of it this way as human evolution.
The team performed 98 different scenarios with three types of robots. Twelve people watched about 200 videos of interconnected robots, then had to compare physical and social goals. “In most cases, their appearance was consistent with what people thought about what was happening on each species,” the researchers said.
The researchers expect the results to be a “benchmark” that allows others to work in a similar way. Next, they plan to create a more complex environment with 3D assistants that allow multiple types of connections. Its goal is not just to teach robots how to better communicate with humans, but to “dig deeper into human nature,” said lead author Andrei Barbu. “Can we do a proper test for you to recognize human interactions? Maybe there’s a way to train people to recognize these interactions and improve their skills.”
All sales supported by Engadget are selected by our writing team, independent of our parent company. Some of our articles include links to links. If you purchase something through one of these links, we may be able to find a support service for you.