By Alex Hupe
Mappix founder, Dr. Chris Johnson, and I sit down to talk about his current endeavors, Mappix, and more.
Could you talk about your current work with unmanned robotic vehicles?
In summer of 2019, I won a sizable grant from the Army to serve as the simulation lead for a large 3-year project in collaboration with Carnegie Mellon, MIT, Penn State, UT-Austin and Texas A&M. We’re building advanced sensor systems for unmanned flying and driving military vehicles, and simulation is a critical component to training the robots. Similar to the way in which we use human-in-the-loop simulation to train pilots, we use simulation for robot-in-the-loop training as well.
This is a new field of research in which we use simulation to expose robotic vehicles to similarly challenging scenarios in make-believe threat environments, like putting a student pilot in JFK Jr’s accident scenario. In fact, Boeing, Bell, Airbus, Uber, Waymo, Argo AI and other leaders in the autonomous vehicle industry use simulation extensively to support machine learning.
Ok, that sounds really interesting. Could you talk about how the development of Mappix relates to all of this?
Great question. During phase 1 of our Army research in fall of 2019, my grad student and I used drones to capture over 10,000 20MP still photos (~111Gb) of an area of rugged terrain owned by Carnegie Mellon near Pittsburgh where they test their robotic vehicles. We then processed the photos using photogrammetry to build a 3D model of the terrain for use in simulation. In fact, drones are mapping the world in high definition, creating thousands of 3D models of natural and man-made structures daily. Of course we had to clean up the photogrammetric model and add physics properties and terrain meshes. Then we built digital twins of the flying and driving robotic vehicles and their sensor systems, and we use the simulation platform to optimize the autonomy software and to train the perception system through the generation of synthetic sensor data.
That being said, my vision for Mappix is to create a robust, world-wide, image repository where people can buy, sell and share geotagged drone imagery for things like creating virtual reality models. To do this, we have leveraged the free flight-training resources that I’ve developed over the years at the University of Wisconsin to build a community of licensed drone pilots. Our free drone certification course has been available for several months, and this month we will be releasing Mappix, which is our user-friendly application for managing geo-tagged drone imagery. Mappix is a “map for your pics”, allowing users to easily manage, discover and share imagery that can be used to develop 3D models of the world. In fact, I wanted to call Mappix “PlaceBook” because the platform is similar to Facebook, but for pictures of places, not faces. We’re building a social network of drone pilots and aerial imagery enthusiasts to buy, sell and share drone imagery, which will help simulation developers like me access photogrammetric imagery to create 3D models for use in simulation, games and virtual reality.
Your free drone training is powered by Pilot Training System. What is the connection, and why are you offering it for free?
Pilot Training System is the company I started while writing my dissertation in 2011 to commercialize my weather simulation system mentioned earlier, and when I started the company, I also wanted to offer free ground school training to solve my own problem as a flight instructor. I wanted my own eLearning system to help my students learn how to fly asynchronously, so I built Pilot Training System. I used the system to build manned and unmanned flight training courses at the University of Wisconsin, which became very popular, and I offer the system for free because I don’t think anybody should have to pay for ground school to learn how to fly.