MIT researchers have developed a method that can be used to automatically design and speed up a custom hardware robot or “brain”.
MIT researchers want to delay robots’ decision-making by developing a method called robomorphic computing Reduce. In this method, the delay between the “mind” and the movements is reduced by creating custom chips according to the physical structure and uses of the robot. The researchers believe that this system can improve the performance of robots in a variety of areas, including robots used in the care of Covid-19 patients.
Researchers say that three important factors are involved in the performance of a robot: perception, mapping and location, planning for decision making. Robomorphic Computing wants to speed up this process by creating custom hardware that performs computing tasks more efficiently.
Using this method, the user first enters the physical parameters of the robot and the system converts them into mathematical matrices. The system then creates a hardware architecture to perform the calculations on the created matrices. As a result, with the help of this system, a chip can be designed that has been customized according to the processing needs of the robot.
The researchers say that the programmed chip uses this method 8 times faster than the CPU and 86 times faster than the GPU, and they hope to create a custom motion-planning chip for all robots. In the future, they want to fully automate the robomorphic computing process so that the user will only be able to create a hardware description by dragging and dropping the robot parameters.