Presenting AirSim


Interesting week! We added a new group member this week, Daniel. This is an intense project, and we can certainly use the help. It was good timing, because we had the Project Plan due this week, and hadn’t parceled out tasks yet.

The other big news this week is that we picked a framework. Initially, I had really been pushing for a 2D simulation. I have a family, work full time, and go to school, so I wanted something manageable and achievable. However, the more research I did, the more convinced I became of the viability of 3D simulations as a project plan. There’s a great series on Youtube and documented online called Python Plays GTA. It’s available here:

https://pythonprogramming.net/game-frames-open-cv-python-plays-gta-v/

Basically, you teach a machine to drive a car in Grand Theft Auto V. Sounds daunting, but as it turns out, the logic is mostly handled in three files, none of which is more than 500 lines or so. Totally digestible. There was also the Udacity Self Driving Car Sim, available here:

https://github.com/udacity/self-driving-car-sim

that lets you drive cars around a 3D environment. Similarly, projects built with that framework tended to consist of 3 important files.

This all preceded an even better find by one of my group mates, Corey. Corey found AirSim:

https://microsoft.github.io/AirSim/

AirSim is a 3D environment that you can navigate around in a drone. It was built to train algorithms for flying drones. A car feature was then added, allowing one to pilot a car around the environment. Basically, it’s the best parts of the environments provided by the other two, packaged up with a host of tools and features for the itinerant developer.

This week, I’ll be getting my AirSim environment set up, and working with a depth camera. Wish me luck!

Print Friendly, PDF & Email

Leave a Reply

Your email address will not be published.