Expert Interview Series: Sasawat Prankprakma of maavumich.org on Lessons From The International Aerial Robotics Competition
Sasawat Prankprakma is the Controls Subteam Lead of Michigan Autonomous Aerial Vehicles, one of the top teams participating in the IARC Challenge.
One of the main reasons for MAAV’s existence is to provide practical, hands-on experience for students by competing in the International Aerial Robotics Competition (IARC). How did you get started working with students, initially?
We are a student team founded by students for students. I joined MAAV Freshman Year at University of Michigan.
What are some of the main benefits of getting students started operating drones early in life?
I don’t really know the main benefits of working with drones in particular, but MAAV lets undergraduate students participate in groundbreaking research driven by the IARC competition. Through MAAV, students are exposed to both the theoretical concepts behind computer vision, artificial intelligence, quadrotor dynamics, and control systems, and the nitty-gritty practical implementation of those concepts in writing software, designing and populating circuit boards, and designing and constructing airframe components. MAAV is proud to do most of the work to compete in IARC in-house (as opposed to modifying off-the-shelf solutions), which not only allows us more control over how the vehicle performs, but provides opportunities for students to learn many different skills and have many different, useful experiences.
The IARC was started way back in 1991. When did MAAV first get involved with the competition? How much has it grown since you first started?
MAAV started competing during the 6th IARC Challenge. I joined during the second year of the 7th IARC Challenge. As of now, no team has successfully completed the 7th challenge. Since MAAV formed, an additional Asia-Pacific venue in Beijing was opened up for the challenge make it easier for Asian and Australian teams to compete.
One of the things that’s unique about the IARC is that the drones are all autonomous – meaning no one’s operating them. Can you talk a bit about the history of self-piloting drones, and how far the field has come since MAAV’s been involved?
While unmanned aerial vehicles have been around for decades, and autonomous waypoint-based outdoor flight has been available at a hobbyist level, drones still have trouble solving strategy and pathfinding problems autonomously, due to a lack of computational resources. When MAAV was first created, the 6th IARC challenge was conceived as a challenge that no then-currently available drone was capable of completing. Advancements in many technologies such as LIDAR based SLAM (Simultaneous Localization and Mapping) lead to the successful completing of the 6th IARC challenge. The 7th IARC challenge forces teams to autonomously interact with ground robots, and navigate without using LIDAR SLAM for localization.
Truly independent aerial vehicles are self-stabilizing, self-navigating, and able to interact with objects on the ground. What are some of the challenges of creating a self-stabilizing drone? Are there any stabilizing devices you’d particularly recommend, either commercially or otherwise?
Self-stabilization is largely a solved problem, though there are still improvements to be had, especially in regards to performance during takeoff and landing. Many off the shelf solutions such as the DJI Naza work very well nearly out of the box.
Localization (finding out where you are) and navigation (figuring out where you are going) are the biggest hurdles to overcome in IARC. For example, the most well developed solutions for localization — GPS and LIDAR SLAM — are both prohibited in the 7th IARC challenge to encourage teams to use computer vision based solutions for localization.
Similarly, how far has the self-navigating apparati come in the past few years? Has GPS gotten more elaborate and detailed, or has it been the same throughout?
We are prohibited from using GPS or LIDAR SLAM for localization. This is reflective of reality, because there are many situations where one might need to fly in an outdoor (no LIDAR SLAM), GPS-denied environment.
There are many off-the-shelf GPS navigation solutions that work very well from companies like DJI and PixHawk.
LIDAR SLAM is much better understood than vision-based localization, however there aren’t any general purpose, off the shelf, out of the box working solutions for LIDAR SLAM like there are for GPS.
In regards to self-navigation, especially via ground objects, what kind of software or programming language does a drone need to run, to be able to differentiate objects on the ground? What are some potential applications of this feature?
We use OpenCV for computer vision. OpenCV is an open-source computer vision library that provides many useful tools for implementing computer vision algorithms and is pretty much an industry standard for computer vision.
The IARC is organized around “missions” – currently in its seventh iteration – where the teams have to conquer a particular challenge. What was the first mission MAAV got involved with? How did it go?
MAAV started competing in Mission 6. The mission was to covertly steal a flash drive from some room in a building with no prior knowledge of the building and deposit a decoy in its place. We were one of the top competitors every year, however, we were not the team that beat the mission.
The current mission involves herding 10 ground objects from the air – 5 of which are red, and 5 of which are green. What are some of the complications, regarding differentiating between different colors from the air? What are some potential uses for this technology, especially in non-military sectors?
There could be some complications with recognizing different colors of objects if the lighting is inconsistent or if parts of the ground happen to be similar colors to the ground objects. Recognizing ground objects is important for a wide variety of military and non-military applications, for example detecting specific objects such as birds, children, or pets would allow delivery drones to safely complete their mission.
How has being involved with the IARC helped to motivate your students? How can the hands-on approach towards STEM applications help students gain a lifelong appreciation for learning and STEM topics? What might be some benefits of this fondness for learning?
We all chose to be involved with MAAV and IARC because it allows us, mainly undergrads, to involve ourselves in bleeding-edge research about drones and robotics. Most other competitions that undergrads get to participate in mainly involve solving problems that are already well understood, but IARC involves solving a problem that is not possible with technology currently available. The hands on approach builds skills valuable for employers and provides a greater appreciation for real world problem solving.
Want to learn more about the future of drones? Contact us today!