An improved chip helps bee-sized drones navigate through space

Researchers at the Massachusetts Institute of Technology, who last year developed a computer chip that helps drones the size of bees navigate space, have now managed to upgrade their chip even further, both in terms of size and energy consumption.

A technical design, which allows for the use of low energy, will allow devices the size of a bee to determine their direction during flight.

Researchers at the Massachusetts Institute of Technology, who last year developed a computer chip that helps drones the size of a bee navigate through space, have now managed to upgrade their chip even further, both in terms of size and energy consumption. The researchers, from the Department of Electrical Engineering and Computer Science, built an entire chip from the base to the processing, focusing on reducing energy consumption and reducing its physical size, improvements that lead to increasing the speed of information processing. The innovative computer chip, with a size of 20 square millimeters, consumes energy at the rate of only 24 milliwatts, a value equal to one thousandth of the energy needed to operate an electric light bulb. Using such a small amount of energy allows the chip to process, in real time, images from a camera at a rate of 171 frames per second, along with making continuous measurements, two features that allow the chip to determine where it is in space. The researchers say that the chip could be embedded in "nano-drones" the size of a fingernail, with the goal of allowing vehicles to navigate, especially in remote or inaccessible places where data from navigation satellites is not available. The chip could also be used in any device or tiny robot that is required to navigate for long periods of time when it has limited energy.

"I can imagine using this chip in the field of robotics based on low energy consumption, such as aircraft with flapping wings the size of a fingernail, or aircraft that weigh less than air, such as air balloons, which are able to operate for whole months without changing their battery", said the lead researcher. "Or imagine medical devices, a kind of capsule that you can swallow, and which can navigate intelligently using an extremely tiny battery that prevents overheating inside the body. The chips we are developing can help with these applications."

In recent years, multiple research groups have developed miniaturized drones so small they can be held in the palm of your hand. The scientists imagine tiny navigational vehicles that could fly around and take pictures of our surroundings, kind of like mosquito-sized photographers or mappers, before coming back and landing on the palm of our hand. At the same time, a drone the size of a palm can carry a limited amount of energy from a battery, energy that is mostly used to run the aircraft engine, leaving only a small amount of energy for other essential operations, such as navigation, and in particular, situation estimation, or giving the robot the ability to determine the Its exact location in space.

"In the field of normal robotics, we take ready-made computers and add algorithms to them, since we usually do not consider the problem of the available energy limit," says the lead researcher. "But in any project that requires us to minimize the applications that consume little energy, we must treat the programming challenges in a completely different way."

In their previous research, the scientists began to address such problems by combining logarithms and hardware on a single chip. Their initial design was implemented on a field-programmable gate array (FPGA), a commercial hardware infrastructure that can be embedded within a given application.

The chip was able to perform a situation estimate using 2 watts, compared to standard and larger drones that use 10 to 30 watts to perform the same task. And yet, this energy consumption of the chip was greater than the total amount of energy required, usually, for miniaturized drones, which the researchers estimate at a rate of about 100 milliwatts.

In order to reduce the energy consumption of the chip, the research group decided to minimize the amount of data - in the form of images from a camera and continuous measurements - stored in the chip at any given moment. The design also allowed for an upgrade of the way this data is transferred across the chip. "All the images that we would otherwise store temporarily on the chip, we actually shrink and thus use less memory," explains the researcher. The researchers also reduced unnecessary operations, such as performing calculations with zero that lead to a zero result – the researchers found a way to omit these computational steps involving zeros. "This method allowed us to avoid processing and storing all these null results, so we could reduce a large amount of unnecessary storage and calculation cycles, a result that led to a reduction in the physical size of the chip and an increase in its processing speed," concludes the lead researcher. Thanks to their innovative design, the researchers were able to reduce the chip's memory from a previous level of 2 megabytes, to a level of only 0.8 megabytes. In addition, the researchers were able to reduce the energy consumption from 2 watts to 24 milliwatts. The research team plans to demonstrate its chip's performance by embedding it inside a miniaturized race car, and eventually inside mini-drones.

The news about the study