Virtual cockpits?

[ShadesOfGray]

Hey,
So, in the coming year, a university development team that I'm working with is going to build a UAV of some sort. Due to our application, the UAV either needs to be fully autonomous, or it at least has to supply us with some kind of virtual cockpit so that we can fly the airplane without visual cues [observing the attitude of the airplane from the ground]. I've been looking at some of the more popular autopilots (MicroPilot, Kestrel), and they're pretty expensive.

So I'm wondering whether anyone sells the instrumentation and software for a virtual cockpit, e.g. some gimbals and accelerometers or whatever for onboard the airplane, and the software that translates that into a virtual cockpit complete with attitude indicator, altitude, airspeed indicator, etc and hopefully some sort of GPS positioning. If we had something like that, we could feasibly fly the airplane without visual cues from the ground to somewhere where it could be landed visually.

Any thoughts?

[clolson]

I'll jump in here with a few thoughts.

First of all, sending back a remote video image is probably going to be your quickest and simplest solution. If you add one of the available text overlay systems that can superimpose gps data on the video, that's even better.

Addressing the popular autopilots: if you add up the time it would take for your team to develop something of similar quality and capability and robustness, the available autopilots start looking like a real bargain. That said, a lot of folks (myself included) have a lot more time than money. (Maybe it would be more accurate to say that I don't have enough time or enough money?) :-) Even a pretty skilled team of engineers could literally spend years putting together an autopilot and tuning and refining it and making it simple and robust. There is an awful lot of skill and knowledge and experience that needs to go into these things to make them work well.

But since you are specifically asking about virtual cockpits, let me put in a plug for two things I think are very cool.

1. FlightGear. FlightGear is a free open-source flight simulator. You can drive the visuals from an external data source. If you put an IMU & GPS onboard your aircraft and send the data to the ground via some sort of wireless comm link (i.e. a radio modem) then you can drive FlightGear from the live data and watch a live virtual 3d synthetic view of the world. You can populate your flying space with aerial imagery, custom objects, whatever you want to add to make it look more realistic. I've posted this movie before, but here is an example of what you can do. This movie shows a video camera view side by side with the virtual flightgear view:

http://youtube.com/watch?v=SAMGnK9ztdA

Notice the row of virtual flight instruments across the bottom of the virtual view. In addition, with flightgear you can create (or use existing) 3d cockpits. You can overlay a HUD, you can drop "bread crumbs" in the sky to visualize your flight path, you can see an "external" view of your aircraft with animated control surfaces.

If you have a net connection at your flying site, you can connect your copy of FlightGear to the flightgear multiplayer system. If your copy of flightgear is being driven from real, live UAV flight data, now you've just inserted a real live aircraft into the FlightGear multiplayer system. People around the world can watch you fly in a live 3d world. In addition, your aircraft will show up on the flightgear google mapping server so anyone in the world with a web browser can follow your flight in real time.

There is a lot of cool and fun things you can do with FlightGear, and I'm just scratching the surface here.

2. I've been working with a small company in LA to adapt their glass cockpit display software for UAV use. They have a really neat version of their product that is modeled to resemble a cutting edge Avidyne glass cockpit display like you would find in a Cirrus aircraft. I've been working on an animated flight director (vbars), heading bug, altitude bug, vertical velocity bug, and control surface position indicators.

If you have a multi-stage PID control loop, you can visualize the input/output of each stage and quite clearly see what needs to be adjusted and where. I've been using it to tune up the autopilot on my flying wing this past week and it's proving to be extremely useful.

For instance, my autopilot uses altitude error to command a target vertical velocity. The vertical velocity error commands a target pitch angle. The target pitch angle commands an elevator deflection. So in this virtual glass cockpit display I see the altitude tape with an altitude bug at the target altitude. I see a vertical speed indicator with the vertical speed bug at the desired rate of climb. Then I see the flight director vbars move to the target pitch angle, and finally I see the elevator surface deflecting to try to achieve that target pitch. I can see the current state of the aircraft along with all the things the autopilot is trying to do. If there are deficiencies in my gains, it's really easy to see where they show up and in which stage. Plus it's really awesome to see this live with all the pieces working together dynamically.

In addition, if you capture your flight data, you can replay it for a much more detailed analysis back in the lab. I've got an early screen shot of the software here:

http://sourceforge.net/project/screenshots.php?group_id=209804

(Oh, I forgot to mention this software has a "nav display" that can show your route "track up" along with real world aviation navaids that may be in the vicinity ... it's modeled after a 747-400 glass cockpit nav display.)

One more thing I should mention. If you plan to control the aircraft from the virtual display, it is critical that you have a high bandwidth, low latency wireless data connection. This is easier said than done. For instance, my maxtream radio modem seems to dynamically scale it's available bandwidth based on distance and signal quality. It's really easy for the available bandwidth to drop below the amount of data I'm trying to send. When this happens, the sending buffers fill up and overrun, and data gets lost or delayed ... I see up to 1.5 seconds delays. That makes it hard to fly from the virtual display. I'm sure this could be improved with maxstream configuration tweaks, and carefully trimming down what data is sent to the ground. But it's an issue you will likely encounter which is why I suggested a video link at the start of this post.

Hope that helps,

Curt.

[ShadesOfGray]

Wow. Thanks for a very helpful reply.

I really like the sound of #1. That sounds very possible. There will be cameras onboard the aircraft, but for obvious reasons, I'd prefer them to be a second reference for the pilot, rather than the sole way that the operator pilots the airplane.

Is IMU = Inertia Measurement Unit? I'm not very familiar with those sorts of things.

FlightGear looks like the way to go for us. I'm actually a licensed airplane pilot with an instrument certification, so flying by reference to virtual instruments such as you showed in your video shouldn't be too challenging.

However, looking over their website, they don't seem to talk alot about using their simulator software in cohort with IMU and GPS to operate a real (ok, scale) airplane. Where can I get more information about that? What would be a good IMU or GPS to use? What setup did you have onboard your RC airplane when you made that video?

Thanks!

[clolson]

However, looking over their [FlightGear] website, they don't seem to talk alot about using their simulator software in cohort with IMU and GPS to operate a real (ok, scale) airplane. Where can I get more information about that? What would be a good IMU or GPS to use? What setup did you have onboard your RC airplane when you made that video?


FlightGear has a standard "C" struct that you can fill in and send over via a UDP (network) socket. For my project I wrote some custom interface software to read in the serial data from the aircraft and reformat it and send it over to FlightGear.

When the video was made we were running a MicroBotics MIDG-II that had been loaned to us. This is a great little integrated IMU/GPS, but it is relatively pricey at $6750/ea.

Many people here are using IR sensor based autopilots which can work really well, or pretty well in most situations most people would want to fly in.

That said we should start a thread to weigh the pros and cons of the available IMU's on the market right now. I say this because I'm thinking about starting to shop for something. I've found a couple good looking options (like the MIDG-II) but they are really expensive and out of my reach. Is there anything good in the $500-1000 range ... that's where I could start trying to talk someone into buying me one to test out. ;-)

Does anyone out there have an "inexpensive" IMU that they've used and that they can recommend?

Curt.

[ShadesOfGray]

Hmm, OK. I was kind of hoping that there would be a set of components that I could put together really easily that would give me the virtual cockpit; if the IMU/GPS is going to cost $6,000+ anyway, then I might as well get a Procereus Kestrel and have fully autonomous flight with their virtual cockpit to use if I want to fly via remote control.

My design group is tasked with designing the structure of the airplane from the ground up, and it has some interesting/novel design parameters of it's own, so I'm kind of trying to keep the electronics design and the time involved there as minimal and simple as possible. Having to write our own interfacing software would be a lot to ask on top of the other aspects of the design that have to be determined.

[macboffin]

Hmm, OK. I was kind of hoping that there would be a set of components that I could put together really easily that would give me the virtual cockpit; if the IMU/GPS is going to cost $6,000+ anyway, then I might as well get a Procereus Kestrel and have fully autonomous flight with their virtual cockpit to use if I want to fly via remote control.

My design group is tasked with designing the structure of the airplane from the ground up, and it has some interesting/novel design parameters of it's own, so I'm kind of trying to keep the electronics design and the time involved there as minimal and simple as possible. Having to write our own interfacing software would be a lot to ask on top of the other aspects of the design that have to be determined. How do you propose to do your final flight testing without the FAA dropping on you like a ton of bricks? They watch this forum as well you know!

[CenTexFlyer]

Uni's somehow seem to get a "pass" on that one Mac..... I dunno how either!

[jetblackaircra]

I'm a bit behind here, but just wondering when you would need your system to be up and running. My company and our partner company are developing our system which we hope will be very affordable and do everything you mentioned. We most likely won't have it available for several months to a year though... Too late for you?

[clolson]

Uni's somehow seem to get a "pass" on that one Mac..... I dunno how either!

There are a few exceptions and a few groups that really do an excellent job building working UAV's, but I suspect most university groups only fly a few seconds at a time between rebuilds. :-)

Curt.