Operating in real or virtual environments, robots are widely used for tasks ranging from carrying out orders in residential and commercial settings to precision operations in manufacturing and healthcare. One of the critical requirements for effective autonomous robot operation is collision avoidance - the ability to navigate its surroundings without crashing into objects. Now, a method developed by researchers at the Massachusetts Institute of Technology (MIT) offers a rapid and accurate way of verifying that a robot's trajectory will not result in a collision.
Older 'safety check' methods have had their challenges. While proving effective in some cases, these methods sometimes falsely certify a trajectory as safe, leading to potential collisions. Those that avoid false positives often operate slowly, unsuitable for real-world scenarios due to the need for quick decisions and actions.
The newly developed system, however, is touted as having 100% accuracy in verifying that a planned robot motion will not collide with any object (assuming the accuracy of the robot and environment model). It's so precise that it can differentiate between trajectories that differ by mere millimeters and usually comes up with proof within seconds. The method simplifies understanding since its mathematical proof can also be checked effortlessly with basic math.
The research team has engineered the technique using an algorithmic method known as sum-of-squares programming, adapted for efficient safety check functionality. This method helps generalize to a wide range of complex motions. It is particularly useful for robots operating within cramped spaces crowded with objects or where collision might cause injury (e.g., robots assisting with food preparation or healthcare).
The technique utilizes the idea of a hyperplane - a mathematical concept where in practical terms, a piece of paper can be visualized separating the robot and objects in the environment, ensuring collision avoidance. As the robot moves, a corresponding hyperplane function modifies to keep the trajectory safe - a continuous versus point-in-time safety check.
By using sum-of-square programming, the team has effectively converted a static problem into a dynamic function, which fine-tunes positioning of the hyperplane corresponding to each point in the planned trajectory for collision-free movements. In essence, the optimization program identifies a group of collision-free hyperplanes.
The researchers have proven this optimization efficient and precise, with the function containing squared values and always positive. This positive nature is an indicator of collision-free trajectory, making verification by humans simpler.
After testing the technique in simulations involving both one-arm and two-arm robots, the process was found to be considerably faster than alternative methods. Although the approach isn't as ultra-fast to be directly implemented in robot motion planning loops that need microsecond decisions, the team is constantly enhancing the process to reach there.
For instance, optimization solvers can quicken the process. Another viable step is to rule out situations not requiring safety checks, like when the robot is at a safe distance from potentially collidable objects. As work progresses, the application of this method would soon allow real-time, accurate safety checks for robots across industries, thereby reducing accidents and enhancing operational efficiency.
Disclaimer: The above article was written with the assistance of AI. The original sources can be found on ScienceDaily.
