Even with all the technological advancements in recent years, autonomous systems have never been able to keep up with top-level human racing drone pilots. However, it looks like that gap has been closed with Swift – an autonomous system developed by the University of Zurich’s Robotics and Perception Group.
Previous research projects have come close, but they relied on optical motion capture settings in a tightly controlled environment. In contrast, Swift is completely independent of remote inputs and utilizes only an onboard computer, IMU, and camera for real-time for navigation and control. It does however require a pretrained machine learning model for the specific track, which maps the drone’s estimated position/velocity/orientation directly to control inputs. The details of how the system works is well explained in the video after the break.
The paper linked above contains a few more interesting details. Swift was able to win 60% of the time, and it’s lap times were significantly more consistent than those of the human pilots. While human pilots were often faster on certain sections of the course, Swift was faster overall. It picked more efficient trajectories over multiple gates, where the human pilots seemed to plan one gate in advance at most. On the other hand human pilots could recover quickly from a minor crash, where Swift did not include crash recovery.
The final results are impressive, especially given that all the processing and sensing comes from the drone. However, it still requires a well mapped track, so a human pilot should still come out on top given limited information about a new track. It would also be interesting to see how it handles large courses with gates that are much further apart.
Professional drone racing is now an elite sport, with all the high-end tech, coverage, and equipment that goes along with it. If you’re just practicing with tiny drones in your home though, you might not be so well equipped. You might want to build something like this tiny FPV drone racing gate from [ProfessorBoots] to help keep track of laptimes while you’re training.
The build uses ultrasonic range sensors to detect when an object passes through the gate. The gate itself consists of a ring of addressable LEDs in strip form. The gate switches from green to red as a visual indicator of a drone passing through the gate. There’s also a small 2.4-inch touch screen that displays laptimes and enables the gate to be configured quickly and easily. The gate also serves up a webpage on the local network for viewing laptimes in a browser.
It does bear noting that at this stage, it’s primarily a practice tool. The gate doesn’t currently work for proper competitions, as it has no way of determining which drone might be flying through the gate at any one time.
Drone racing is nifty as heck, and a need all races share is a way to track lap times. One way to do it is to use transponders attached to each racer, and use a receiver unit of some kind to clock them as they pass by. People have rolled their own transponder designs with some success, but the next step is ditching add-on transponders entirely, and that’s exactly what the Delta 5 Race Timer project does.
The open-sourced design has a clever approach. In drone racing, each aircraft is remotely piloted over a wireless video link. Since every drone in a race already requires a video transmitter and its own channel on which to broadcast, the idea is to use the video signal as the transponder. As a result, no external hardware needs to be added to the aircraft. The tradeoff is that using the video signal in this way is trickier than a purpose-made transponder, but the hardware to do it is economical, accessible, and the design is well documented on GitHub.
The hardware consists of RX508 RX5808 video receiver PCBs modified slightly to enable them to communicate over SPI. Each RX508 RX5808 is attached to its own Arduino, which takes care of low-level communications. The Arduinos are themselves connected to a Raspberry Pi over I2C, allowing the Pi high-level control over the receivers while it serves up a web-enabled user interface. As a bonus, the Pi can do much more than simply act as a fancy stopwatch. The races themselves can be entirely organized and run through the web interface. The system is useful enough that other projects using its framework have popped up, such as the RotorHazard project by [PropWashed] which uses the same hardware design.
While rolling one’s own transponders is a good solution for getting your race on, using the video transmission signal to avoid transponders entirely is super clever. The fact that it can be done with inexpensive, off the shelf hardware is just icing on the cake.
Drone racing comes in different shapes and sizes, and some multirotor racers can be very small indeed. Racing means having gates to fly though, and here’s a clever DIY design by [Qgel] that uses a small 3D printed part and a segment of printer filament as the components for small-scale drone racing gates.
The base is 3D printed as a single piece and is not fussy about tolerances, meanwhile the gate itself is formed from a segment of printer filament. Size is easily adjusted, they disassemble readily, are cheap to produce, and take up very little space. In short, perfect for its intended purpose.
Arguably, drones are one of the next big things that will revolutionize many industries. We’ve already seen them portrayed in many movies and TV shows in not-so-distant futures, and what with Amazon planning drone deliveries, we can’t imagine it’ll be long before they are a common sight flying around cities.
With a small group of people pushing the envelope and inventing every day.
Not for long though. DRL is making a huge push to turn this into a mainstream sport, and we gotta admit — we don’t mind. After all, this is like pod-racing on crack. Just take a look at the following promo video for their course the Gates of Hell: the Dream Takes Form.
Last weekend was Sparklecon, the premier meetup in Southern California of dorks dorking around, fire, electricity, welding, and general mischief. Just imagine a party of a hundred or so like-minded individuals at a hackerspace. Now imagine the entire party is the after party. That’s a pretty good idea of what happened.
The event was held at the 23b shop in Fullerton, a true hackerspace tucked away in a small industrial park. The people at 23b are using their location to their advantage: no one in the neighborhood really cares what happens after 5pm on a Friday. This allows for some very loud, very bright, and very dangerous hijinks.
There was something for everyone at Sparklecon, including:
Electric Pickle. Take a stick welder, and put a few hundred amps through a pickle. First, the pickle turns into a sodium light. Then, it turns into a carbon arc light. Best done after dark.
FPV drone racing. Flying around and crashing into trees in an abandoned lot. FPV from a few quads were projected onto the side of a building
Live music! Analog synths and Game Boys!
Tesla coils! This was a 300 amp monster, and completely analog. The spark gap was impressive by itself, but it gets really cool when you steal a fluorescent light from a fixture and stand 20 feet away from the Tesla coil.
Hammer Jenga! Cut some 2x4s up and make a tower of Jenga. Get a hammer, some colorful commentators, a dozen people, and make some competition brackets. Hackaday’s own [Jasmine] was the first champion of the night.
Sparklebot Death Battle! It’s like BattleBots, only things break more often and we don’t have [Bil Dwyer].
Hebocon! Battling robots, but much crappier than the Sparklebot Death Battle. These robots broke more often.
Analog synths provided the tunes
The Sparklebot Death Battle ring
Tesla Coils and Spark Gaps
A Hebocon bot, using a mouse trap as a weapon
A lady tribble, vibrating her way across the Hebocon ring
The basic premise of Hammer Jenga
Art was made out of the spare parts left over from the Hebocon build-off. This robot is named Art
The main event was, of course, Sparklecon’s own version of Battlebots. There were only four competitors the entire night, but the competition was fierce.
Three of the bots were wedge designs, in keeping with the ramp-ification of battling robots. The lone exception to this was [Charlie]’s Slow Bot, a cube design equipped with a spinning steel blade. The blade moves fast, but Slow Bot doesn’t. It’s a purely defensive design, meant to destroy bots trying for an easy kill. The test video of Slow Bot can be seen here:
The first fight of Slow Bot did not live up to the hype, unfortunately. After Slow Bot’s primary weapon got up to speed, the opposing bot moved in for the kill. The bolts on Slow Bot‘s blade sheared, ending the match, and leaving five or six people looking around the 23b shop for M5 bolts, or some larger bolts and a tap.
Is it all hilarously unsafe? Well, there were some plexiglas shields in front of the crowd, and most people viewed the fights on the projector beaming against the wall, anyway.
Is it worth it to go to Sparklecon? If you like dangerous experiments, soldering wires directly onto AA batteries, fire, electricity, electromagnetic fields, broken robots, and hanging out by a fire, yes. It’s a party at a proper hackerspace, making it the best kind of party ever. If history repeats itself, there will also be an afterparty at 23b following the LayerOne conference in May.