Flying a quadcopter or other drone can be pretty exciting, especially when using the video signal to do the flying. It’s almost like a real-life video game or flight simulator in a way, except the aircraft is physically real. To bring this experience even closer to the reality of flying, [Kevin] implemented stereo vision on his quadcopter which also adds an impressive amount of functionality to his drone.
While he doesn’t use this particular setup for drone racing or virtual reality, there are some other interesting things that [Kevin] is able to do with it. The cameras, both ESP32 camera modules, can make use of their combined stereo vision capability to determine distances to objects. By leveraging cloud computing services from Amazon to offload some of the processing demands, the quadcopter is able to recognize faces and keep the drone flying at a fixed distance from that face without needing power-hungry computing onboard.
There are a lot of other abilities that this drone unlocks by offloading its resource-hungry tasks to the cloud. It can be flown by using a smartphone or tablet, and has its own web client where its user can observe the facial recognition being performed. Presumably it wouldn’t be too difficult to use this drone for other tasks where having stereoscopic vision is a requirement.
Thanks to [Ilya Mikhelson], a professor at Northwestern University, for this tip about a student’s project.
They say that drummers make the best helicopter pilots, because to master the controls of rotary-wing aircraft, you really need to be able to do something different with each limb and still have all the motions coordinate with each other. The control complexity is due to the mechanical complexity of the swashplate, which translates control inputs into both collective and cyclical changes in the angle of attack of the rotor blades.
As [Tom Stanton] points out in his latest video, a swashplate isn’t always needed. Multicopters dispense with the need for one by differentially controlling four or more motors to provide roll, pitch, and yaw control. But thanks to a doctoral thesis he found, it’s also possible to control a traditional single-rotor helicopter by substituting flexible rotor hinges and precise motor speed control for the swashplate.
You only need to watch the slow-motion videos to see what’s happening: as the motor speed is varied within a single revolution, the tips of the hinged rotor blades lead and lag the main shaft in controlled sections of the cycle. The hinge is angled, which means the angle of attack of each rotor blade changes during each rotation — exactly what the swashplate normally accomplishes. As you can imagine, modulating the speed of a motor within a single revolution when it’s spinning at 3,000 RPM is no mean feat, and [Tom] goes into some detail on that in a follow-up video on his second channel.
If you’ve ever flown or watched anyone fly a racing drone for any length of time, you know that crashes are just part of the game and propellers are consumables. [Adam] knows this all to well, decided to experiment with combining multiple broken propellers into one with a 3D printed hub.
A damaged propeller will often have one blade with no damage, still attached to the hub. [Adam] trimmed the damaged parts of a few broken props, and set about designing a 3D printed hub to attach the loose blades together. The hubs were designed let the individual blades to move, and folding out as the motors spin up, similar to the props on many photography drones.
Once [Adam] had the fit of the hubs dialed in, he mounted a motor on a piece of wood and put the reborn propellers through their paces. A few hubs failed in the process, which allowed [Adam] to identify weak points and optimise the design. This sort of rapid testing is what 3D printing truly excels at, allowing test multiple designs quickly instead of spending hours in CAD trying to foresee all the possible problems.
He then built a test drone from parts he had lying around and proceeded with careful flight testing. The hubs were thicker than standard propellers so it limited [Adams] motor choices to ones with longer shafts. Flight testing went surprisingly well, with a hub only failing after [Adam] changed the battery from a 3 cell to a 4 cell and started with some aerobatics. Although this shows that the new props are not suitable for the high forces from racing or aerobatics/freestyle flying, they could probably work quite well for smoother cruising flights. The hubs could also be improved by adding steel pins into the 3D printed shafts, and some carefully balancing the assembled props.
Drones (and by that we mean actual, self-flying quadcopters) have come a long way. Newer ones have cameras capable of detecting fast moving objects, but aren’t yet capable of getting out of the way of those objects. However, researchers at the University of Zurich have come up with a drone that can not only detect objects coming at them, but can quickly determine that they’re a danger and get out of the way.
The drone has cameras and accompanying algorithms to detect the movement in the span of a couple of milliseconds, rather than the 20-40 milliseconds that regular quad-copters would take to detect the movement. While regular cameras send the entire screens worth of image data to the copter’s processor, the cameras on the University’s drone are event cameras, which use pixels that detect change in light intensity and only they send their data to the processor, while those that don’t stay silent.
Since these event cameras are a new technology, the quadcopter processor required new algorithms to deal with the way the data is sent. After testing and tweaking, the algorithms are fast enough that the ‘copter can determine that an object is coming toward it and move out of the way.
It’s great to see the development of new techniques that will make drones better and more stable for the jobs they will do. It’s also nice that one day, we can fly a drone around without worrying about the neighborhood kids lobbing basketballs at them. While you’re waiting for your quadcopter delivered goods, check out this article on a quadcopter testbed for algorithm development.
We were concerned when we saw [Brick Experiment Channel] test a drone propulsion pod made with Lego. After all, the thrust generated was less than the weight of the assembly. But a few tweaks got enough lift to overcome the assembly weight, as you can see in the video below.
The next step was to build three more pods and add some lightweight avionics and a battery. The first flight was a little dicey because the sensor orientation was off. Then there was some more software tuning before things really got airborne.
Small hobby aircraft and light plastic parts go hand in hand, and a 3D printing pen makes lightweight plastic things without the overhead of CAD work and running a 3D printer. So could a 3D pen create useful plastic bits for small quadcopters? [Michael Niggel] decided to find out by building his drone parts with a 3D pen loaded with ABS plastic. He mostly discovered that the created objects could politely be said to look like they were sketched by a toddler, but that’s not all he learned.
He found that in general creating an object was harder than the marketing materials implied. As soon as the filament exits the pen’s nozzle, the thin little molten line of plastic cools rapidly and does two things: it has a tendency to curl, and loses its desire to stick to things. [Michael] found the whole affair worked much less like ‘drawing in thin air’ and rather more like piping frosting, or caulking.
Nevertheless, [Michael] sought to discover whether a 3D pen could be used to make quick and dirty parts of any use. He created two antenna brackets and one micro quad frame. All three are chaotic messes, but one antenna bracket was perfectly serviceable. The 3D pen was indeed able to create a strangely-shaped part that would have been a nightmare to CAD up. The other antenna part worked, but didn’t do anything a zip tie wouldn’t have done better. The rapid cooling of the plastic from the 3D pen has an advantage: extrusions don’t “droop” like a glob of hot glue does before it hardens.
By now, [Michael] agreed that the best way to create a plastic part of any complexity whatsoever seemed to be to draw sections flat, build them up in layers, then use the pen to weld the pieces together and add bulk. The micro quad frame he made in this way doesn’t look any nicer than the other attempts, but it did hold the parts correctly. Sadly, it would not fly. Once the motors powered up, the arms would twist and the flight controller was unable to compensate for motors that wouldn’t stay straight. This could probably be overcome, but while the end result was dirty it certainly wasn’t quick. The 3D pen’s niche seems restricted to simple, unstressed parts that aren’t permitted to gaze up themselves in a mirror.
When it comes to robotic navigation, the usual approach is to go as technically advanced and “smart” as possible. Yet the most successful lifeforms that we know of follow a completely different approach. With limited senses and cognitive abilities, the success of invertebrates like ants and honeybees lie in cooperation in large numbers. A joint team of researchers from TU Delft, University of Liverpool and Radboud University of Nijmegen, decided to try this approach and experimented with a simple navigation technique to allow a swarm of tiny flying robots to explore an unknown environment.
The drones used were of-the-shelf Crazyflie 2.0 micro quadcopters with add-on boards. Sensors consisted of it’s onboard IMU, simple range finding sensors on a Multi-ranger deck for obstacle detection, and a down pointing optical flow sensor, on a Flow deck, to keep track of the distance travelled. To navigate, the drones used a “swarm gradient bug algorithm” (SGBA). Each drone in has different preferred direction of travel from takeoff. When an obstacle encountered, it follows the contour of the obstacle, and then continues in the preferred direction once the path is clear. When the battery drops to 60%, it returns to a wireless homing beacon. While this technique might not be the most efficient, it has the major advantage of being “lightweight” enough to implement on a cheap microcontroller, an STM32F4 in this case. The full research article is available for free, and is a treasure trove of information.
The main application researchers have in mind is for search and rescue. A swarm of drones can explore an unstable or dangerous area, and identify key areas to focus rescue efforts on. This can drastically reduce wasted time and risk to rescue workers. It is always cool to see complex problems being solved with simple solution, and we are keen to see where things go. Check out the video after the break. Continue reading “Tiny Drones Navigate Like Real Bugs”→