In the course of researching another piece, I came across a picture of a solder hand-launching an RQ-11 drone sorry, Unmanned Aerial Vehicle (UAV) in an article on military connectors. The text mentioned that researchers are working on exploiting an idea from Mother Nature- flapping wings – to develop a Micro AV (MAV) that looks more like a bird, thereby achieving a degree of camouflage.
Enquiring minds want to know… it turns out there's quite a lot of research going on at universities and in the military-industrial complex into Ornithopters and insect-like “bugbots”, with breathless promotional videos on YouTube like this creepy example.
Without getting into the scarier aspects of these beasts, as a recovering engineer I was naturally intrigued by the technology behind it all. Most of the images you can find online seem to be the product of Photoshop and an overheated imagination, but the device in Figure 1 is real. It's the DelFly Micro from Delft University of Technology in Holland – a bugbot that weighs in at a mere 3.07 grams (0.15 oz) and measures 10cm (3.9 inches) from wingtip to wingtip.
Here's how it breaks down:
The wings, which flap at 30Hz, are constructed of mylar foil, carbon and balsawood. The battery is a 30mAh Lithium polymer, good for 3 minutes of flight and range of 50m. The whole thing is controlled by a human pilot, with three controls for the throttle, elevator and rudder.
Impressive enough, but if you want to make the world safe for democracy, you need something more. The next generation DelFly Explorer, created in 2013, is larger at 20g with a 28 cm wingspan, but incorporates a stereo vision system rendering it capable of 9 minutes of autonomous flight. In order to accomplish this, the DelFly Explorer also includes an autopilot with 3-axis accelerometers, a magnetometer and a barometer.
Two processors are used: an Atmel Atmega328P, based around the AVR low power 8-bit core, for autopilot control; and an STM32F405, based around an ARM Cortex-M4 core, for onboard vision processing.
The motive power is provided by a hand-wound BLDC motor with 32 windings around the coils. This was reduced from 37 in earlier models in order to increase the speed (and hence lift) for the available input voltage (3.5V), at the cost of reduced torque.
Processing the stereo images to achieve autonomous flight presented special problems, not least because the flapping of the wings induced distortions in the 128×96 pixel images. The vision software avoids obstacles by analyzing pixel disparities between the left and right part of the stereo image. If the total number of pixel disparities is lower than an empirically determined threshold, the DelFly will continue to fly straight; otherwise it will turn towards the side with the fewest such disparities.
In the interest of computational efficiency some typical stereo vision steps such as un-distortion and image rectification are skipped, resulting in a final program execution time of 90ms.
Some results are shown in Figure 3 where the DelFly was programmed to avoid (solid blue line) or target (dotted red line) vertical poles (black circles) in the test area. The dotted blue line shows a trial where the device hit two poles with its wings.
Development is continuing, although thankfully we're still some way from this nightmare scenario.