VICON VALKYRIE CASE STUDY: BROWN UNIVERSITY ENGINEERING

As use cases move from individual drones to swarms, new problems related to coordination and aerodynamics emerge and compound.

“I’m involved in solving a two-part problem,” explains Anoop Kiran, a PhD student at Brown. “First is the aerodynamics of multi-robot flights, which is the physics of flow, meaning the airflow that’s generated by the propellers of rotors as they rotate on a drone, and the interaction between drones as they fly close together. I’m studying the physics of it, quantitatively modeling it, and coming up with control algorithms which can provide robustness and stability when the swarm scales to a larger formation.”

In particular, Kiran is interested in the effects of ‘downwash’—the volume of air pushed downward by drone propellers as they rotate—on other drones in a formation, and in ground effect – the enhanced lift a drone experiences when it flies close to the ground.

“I hope to inform the scientific community about close proximity flight. We’d like to know exactly how close drones can fly together such that they can lift an object precisely, accurately and have it delivered to a location. Or this could be tied to search and rescue,” says Kiran.

“A lot of researchers have looked at the problem,” he explains, “but literature often essentially ignores the problem itself by focusing on having drones fly farther apart from each other or having some sort of occlusion or effect that might not emphasize down wash directly. So we’re looking at how many drones we could fly close together before they finally lose stability. And to achieve that, accurate position tracking is required. We rely on Vicon to solve that problem, so that we can focus on better control algorithms that can stabilize the drones when the aerodynamics become unsteady.”

“Our main robots are small Crazyflies that fit in the palm of your hand,” says Eric Ewing, a PhD candidate and visiting assistant professor at Brown. “We’ve flown up to 50 of these at a time, and we’ve flown experiments with 32 of them, but we also do things with fewer robots.”

In these experiments Brown’s Vicon motion capture system essentially stands in for GPS, which would be used in an outdoor environment, giving highly accurate information on the exact position of the crazyflies.

“We have 14 Vicon Valkyrie cameras,” says Ewing. “We use Vicon Tracker 3.10 at the moment, which serves all of our needs. We do both rigid body tracking with objects built into Vicon Tracker and custom interfaces for single marker tracking. We send all of the data we get from Vicon Tracker over to a computer that runs ROS (the Robot Operating System) and we send this information to the drones. They either use it as their state estimation or to compare against their own state estimation to measure how far off the ground truth their sensors are reading.”

The lab’s use of single markers is a response to issues with marker occlusion that occur when multiple drones fly in close proximity to each other.

“Instead of having multiple markers, we ended up deciding to have a single marker on each drone,” explains Kiran. “The point of this was that when we have a formation which scales up to multiple drones it’s difficult to have a unique identifier for each of those drones.

“We decided to stick to a single marker, but to focus on having larger markers so that the Vicon camera could view the entire area of the marker. Even though it’s just a single marker, now we have a larger surface area to work with. This means that the infrared light that’s emitted from the camera now bounces off of a larger surface, ensuring that accuracy and precision is maintained and we still get high resolution, low latency data.”

Brown’s Vicon Valkyrie cameras are an important part of the solution.

“I feel very fortunate to have so many cameras that can capture images at 50 frames per second,” says Kiran. “That kind of transmission rate is hard to beat in terms of precision. It offers great high-resolution, low-latency, highly-precise motion tracking, which enables us to capture data real-time and process it, something that’s very hard to find in the market. It’s an essential technology. It makes our job easier in the sense that problems related to localization are solved, so that I can focus on other aspects of research that are more important to me, like control.”

As well as enabling Kiran to focus on the aspects of his projects that are most important to him, Brown’s Vicon system has led to productivity gains. “One of the biggest lessons that the Vicon system has taught me is that real-time data capture makes analysis much easier,” says Kiran. “It ensures that I’m performing analysis in a timely manner rather than pushing it aside for later. It enhances my productivity.

“Aside from that, the wireless capability makes things easier in terms of data transmission. We get real-time feedback from the system, which is something that’s hard to match. Normally data processing takes a whole lot of time, which is something we can cut down thanks to the Vicon system.”

Ewing was impressed by how simple it was to get the lab’s Vicon Valkyrie cameras set up. “We had one of the first Vicon Valkyrie systems installed, and the installation process was incredibly smooth considering how new these cameras were. They just worked out of the box, which is very impressive. They’re perfect,” he says.

 

More importantly, the Vicon Valkyrie system enables Ewing, Kiran and their colleagues to perform experiments they wouldn’t have been able to do otherwise. “Prior to the Valkyrie system we had the older Vantage cameras, but the new cameras really blow them away in terms of field of view and their precision. Now we can cover twice the flight space with the same amount of cameras,” says Ewing.

“Having twice the space has helped the research a lot because we do experiments with up to 50 drones. Working with those in our previous space, there was only so much we could do. But being able to cover twice the space with the same level of precision, the same level of speed and performance, has really made it easier to reliably run larger experiments with more drones.”

Ewing has consulted on the setup of numerous new volumes created to do multi-robot research. “I’ve recommended Vicon to people starting labs multiple times at this point,” he says. “I think the precision that Vicon gives you compared to its competitors is a big benefit. Quadcopters don’t have a lot of sensing onboard, and if we don’t want to add additional sensing, which takes weight and energy, we really need the precision that Vicon gives us.

“With competitor systems, you either need more cameras or the cameras are smaller and less accurate than Vicons. If you want to do multi-robot research, the precision that Vicon gives you is definitely a benefit.”

One of the most significant barriers to entry in robotics education is the complexity involved. Brown’s Vicon system helps lower that barrier by streamlining setup and data analysis, and recently Brown has successfully brought students with no background in engineering or computer science into its lab.

“By far the coolest thing that we’ve done recently is getting these art students who’ve had no robotics experience and letting them come up with art projects that use these robots,” says Ewing. “As engineers, we would never have come up with these types of projects. It really was a joy to work in these cross-discipline groups with engineers, computer science students and arts students. It was really an incredible experience for us.”