Researchers at MIT are working to resolve the considerable difference between a robot’s processing speed (which tends to be comparatively slow) and its physical speed (which is quite fast due to recent advancements in hardware). They are employing a technique known as “robomorphic computing” to achieve this. Developed by Dr. Sabrina Neuman, a graduate of MIT’s Computer Science and Artificial Intelligence (CSAIL) program, this approach involves creating specialized computer chips that provide hardware acceleration for quicker reaction times.

The creation of custom chips designed for a particular function is not a new concept—modern smartphones currently utilize them. However, these chips are gaining prominence as organizations and experts seek to perform more computations directly on devices with limited power and processing capabilities, rather than transmitting data to and from large data centers over networks.

This method centers on developing highly specialized chips that are engineered according to a robot’s physical structure and its intended applications. By considering a robot’s needs regarding its perception of the environment, its ability to map and understand its location, and its planning of movements based on that mapping and necessary actions, researchers can create processing chips that substantially improve the efficiency of the final stage through the addition of hardware acceleration to existing software algorithms.

A common example of hardware acceleration is the graphics processing unit, or GPU. A GPU is a processor specifically engineered for handling graphics-related tasks—such as rendering images and playing videos. GPUs are widely used because most modern computers run applications that demand significant graphics processing power, but custom chips for a variety of functions have become increasingly popular with the emergence of more adaptable and efficient small-batch chip manufacturing processes.

According to MIT News, here is a detailed explanation of how Neuman’s system functions when applied to optimizing a hardware chip design for robot control:

Neuman’s team utilized a field-programmable gate array (FPGA), which represents a middle ground between a completely custom chip and a standard CPU, and achieved notably improved performance compared to the latter. This suggests that even greater performance gains could be realized by actually manufacturing a chip from the ground up.

Accelerating a robot’s response time to its surroundings is not solely about boosting production speed and efficiency—although it will accomplish that. It also contributes to making robots safer to operate in environments where humans are working directly with and alongside them. This remains a major obstacle to the broader adoption of robotics in daily life, indicating that this research could facilitate the realization of a future where humans and robots coexist and collaborate seamlessly.