The miniaturization of drone technology is really fast, small drones can be loaded with other features too. But everything else has shrunk, not the computer chip handing the operations the machine.
Most chips for quadcopters can manage from data that is streamed from sensors as the fly themselves autonomously. Powered by 10 to 30 watts of battery power that will make a smaller bee sized drone laden down.
Brilliant minds at MIT have devised a computer chip that is low powered and can be used for bottlecap sized drones. They call it "Navion" when it was presented at a conference, held at MIT. They showcased it as a technology update on existing drone technology.
What came with the device is a low power algorithm, with mini-hardware, and a special design computer chip (CPU) to operate a drone.
The design process for the chip is by creating the hardware and algorithms, that saves more power. Over the usual of having hardware and algorithms made separately.
When programming robots this is better and scales down everything when both hardware and algorithms are made at the same time.
It just gets better when the chip will process images at 20 frames per second and adjusts its direction with ease. Better yet is the chip will do that work at lower than 2 watts of power. It exceeds the capacities of most drone embedded chips today.
The engineers want to design small and smart drones that can operate on their own. Most likely to be deployed in emergencies as mini-scouts to do search and rescue operations. Another application is for consumer electronic products, but as novelties mostly.
One interesting use of a bottle cap sized drone that is small and portable, to be deployed with a flick of the wrist. Use gestures, and it will capture images, that can be sent wirelessly to your mobile or share in a group.
Available mini-drones can fit a fingertip and very light, uses 1 watt of power to fly up. Other sensors and cams will use 0.5 watts to keep aloft. What is needed now, are computers that are small enough to mount, and work on low power as well.
It became evident that designing a custom chip that is small enough is the answer. Plus, the processing power to smartly fly a robotic drone.
Even with smaller transistors that are vast improvements in speed and efficiency, it is not enough for the specs needed. Building custom hardware from the ground up is the best option.
The computer chips were given a custom algorithm that improved how it oriented itself better and achieving low power consumption. Reducing the gates and optimizing other functions is what made the custom computer chip better. Getting the right process and not using too many process gates was deduced when programming the patterns to follow.
Crucial to using less power than 2-watts is the limitation of 20 frames per second used by the drone camera. With fewer data used by the CPU to process and less than 2-watts, small amounts of data will adapt to the current function. Algorithms for less power and data, yet able to manage all functions is the keystone.
More test on the FPGA computer chip will be done, and learn more about reaching the best performance of the drone computer chip.
Read: New Chip Design Method May Result in Miniature Smart Drones