Robots need to avoid colliding with objects while performing tasks. MIT researchers have developed a technique that proves a robot’s trajectory is collision-free with 100% accuracy in just a few seconds.
Sum-of-Squares Programming
The researchers used sum-of-squares programming, a powerful algorithmic tool, to solve the safety check problem. This technique can discriminate between trajectories that differ by millimeters, making it ideal for robots in crowded spaces with objects.
Certifying Safety
Traditional safety check algorithms can’t predict if a robot will collide with objects in the intermediate seconds. The new technique generates a hyperplane function that moves with the robot, proving the entire trajectory is collision-free. The function is easy to verify and guarantees accuracy, assuming an accurate model of the robot and environment.
If the method proves successful, it could help robots avoid collisions and injuries in various settings such as commercial kitchens and home healthcare. The researchers plan to improve the technique to make it faster for real-world applications.