14 DAY TRIAL // A group of engineers at the Cornell University recently showcased their latest innovation, a Rethink Robotics Baxter machine that has been extensively modified, to the point where it can co-actively learn from humans as it performs various actions.
The research group set up a scenario where the robot acted as a checkout worker who needed to wield a knife. Through several iterations, the machine was able to learn the proper way to hold and handle the knife, without dedicated programming guiding it through every step of the process.
The Baxter robot is capable of planning its own trajectories. However, if the humans it interacts with are not happy with its selection, they are prompted to choose one of three alternative trajectories. Through successive iterations, the robot eventually learns the proper way to accomplish a task.
“We give the robot a lot of flexibility in learning. The robot can learn from corrective human feedback in order to plan its actions that are suitable to the environment and the objects present,” says CU assistant professor of computer science, Ashutosh Saxena, quoted by EurekAlert.