My background is in computer science, specifically computer vision and machine learning, and I have just begun working on a UAV-related project that is in the inception phase.
As part of the project, I need to develop a generic vehicle mission system (i.e. not necessarily for UAVs). The details are a little sketchy at this stage, but as a first step I need to survey the existing autonomous vehicle systems that exist. I am interested in learning what are the typical system components, features, sensors, protocols, APIs, and so forth from implemented systems.
The systems that I am interested in are ones that are relatively mature (i.e. actively developed and currently deployable) with a reasonable number of end-users. They can be free, open source, or commercial. Ideally, the systems are generic (systems for autonomous vehicles in general, not just UAVs) and are modular/extensible, but this is not essential.
What systems would be a good starting point? (e.g. ArduPilot/ArduCopter, Parrot AR.Drone?)
What about for non-UAV systems (UUV, UGV)?
Are there any systems that are non-modality specific (vehicles in general that can be extended to a particular modality, such as UAV)?
Ardupilot is a good starting point. If you haven't flown a rc heli you shouldn't try the quadcopter until you have. I haven't dealt with the Parrot AR drone but I heard its not to hard to fly. The APM board I believe has been used to make a UGV not shure how it turned out. But with APM, if you can program it you can do it (eventually). Could you explain non-modality? I dont understand.
Thanks for the reply Nick. What I meant by "non-modality specific" was that the software system was not designed for a specific type of vehicle, and deployable on a wide range of hardware platforms. That way all vehicle types (planes/helis, boats, subs, cars, etc) can use the underlying system by adding or building modules that extend upon a baseline vehicle framework.
I would expect such a framework would contain all the basic building blocks required by all vehicular systems: navigation, communications, possibly perception (for autonomy).
The closest frameworks I have seen to this is Willow Garage's ROS and Microsoft's RDS. These systems are a little too general, but may serve as a better starting point, whilst using Parrot AR drone and Ardupilot as implementation guides for the additional modules.
You can try looking at the MOOS framework developed by MIT and the US Navy. It's currently used for marine vehicles but can conceivably be expanded to include other platforms as well. MOOS-IVP
Thanks Chris and Adrian; that gives me something to work with.
Has anyone played around with Willow Garage's ROS? If so, how easy is it to extend to a UAV system?