Russian scientists use a quantum computer to accelerate robot movements 30-fold

Using a quantum computer

As reported by Kommersant, Russian researchers have devised a way to accelerate processes involved in robot movement by several dozen times. More specifically, the reported speed-up is 30-fold.

Researchers from the Professor A. N. Gorban Central University Scientific Laboratory of Artificial Intelligence, Data Analysis and Modelling, the Innopolis University AI Institute and other organisations have developed a new technique for faster robot control. This refers to reducing the delay between deciding what position a robot, or one of its components, should take and the point at which that movement actually begins.

The approach is also intended to make movements smoother and more cost-efficient. As a result, the robot would avoid unnecessary motion.

Overall, the work concerns what is known as inverse kinematics. A person does not usually consider precisely how they make a particular movement, such as moving an arm. A robot, however, needs a system that knows exactly how to operate its components in order to perform a given action.

The Russian researchers’ idea is to convert this task into a format suitable for a quantum computer. According to the source, the angles between a robot’s “joints” are encoded as a particular sequence of strings of zeros and ones, while finding the best position becomes a matter of locating the minimum of a quadratic function based on those zeros and ones.

“This format makes it possible to use quantum annealing - a technology implemented in new processors - to find the global minimum in a complex solution space, thereby optimising movement. It is similar to the way numerous human muscles contract and relax so that a hand can precisely pick up a cup of coffee.

Experiments were carried out on a real D-Wave quantum processor. The scientists assessed how chain length - in other words, processor capacity - affects the accuracy of actions and the algorithm’s running time. The findings showed that hybrid quantum-classical algorithms achieved a speed-up of more than 30 times compared with conventional silicon-based methods.”