Researches use AI to improve assistance of robot arms

According to published data, in the U.S., more than one million adults use wheelchair equipped with robot arms to help them carry out everyday tasks. This includes brushing their teeth, dressing, and eating. However, robotic devices currently available can be hard to control. For example, to open a cabinet door or remove a food container from a fridge can take a long time. And, use of a robot to feed an individual is even harder due to fine manipulation required for the task.

In order to improve the assistance of robots, researchers at Stanford have developed a new way to control assistive robotic arms. The method is faster and more intuitive than currently used ones. For trials, the new robot controller allowed individuals to cut tofu efficiently and serve it onto a plate, or dig into marshmallow, scoop it with icing, and dip in sprinkles.

Advances in Robot Arm to help feed Individuals exciting for researchers

“In fact feeding is one of the chosen problems to work on,” said one of the researchers. This is because it is difficult from robotics standpoint, requires precise manipulation and is such a basic task. And, the problem is exciting, the benefits of the robotic arm are right in front of the eyes.

Meanwhile, the research team developed a controller that combines two artificial intelligence algorithms. The first one developed by the team enables two dimensional control on a joystick, without requiring to switch between modes. For example, the algorithm uses contextual cues to determine if the user is reaching a drinking cup or reaching for a doorknob. Following this, as the robot arm is close to its destination, the second algorithm pitches in to allow more precise movements, with sharing of the control between the robot and the human.