MSFS as controller software

[rjfon1]

Morning all.
I'm thinking of commencing a UAV project, and I've been following many threads. Just wondering if I could get a few opinions on this idea.
For controller software, use Microsoft Flight Simulator(MSFS). The object of this would be to:
1) Build the actual UAV vehicle airframe.
2) Build an aircraft within MSFS using normal aircraft building techniques, with the model and profile and .air file as close to the actual as possible to the actual.
2A) A small computer would be installed inside the airframe. This could be a canabalised laptop, as no screen would be required.
3) Using a program file called FSUIPC.dll which is imbedded within MSFS to send and receive output strings from MSFS. It is sort of like networking, with the program on one computer, and the actual aircraft being the other. The aircraft would send altitude, airspeed, tachometer, attitude, bank angle, compass direction to the computer. A GPS module would send Lat., Long., Groundspeed, time to the computer, and this information is converted to the required format for input into FSUIPC and then into MSFS.
4)MSFS already does its own calculations as far as direction, bank angles climb angles, yaw angles, required airspeeds, and sends this info to servo controllers and thus servos.
5)These same strings can be sent via modem link to the ground station running the same program. You see what it sees, and if you want to control it, you do so with a standard joystick. Autopilot control is from the ground with a single switch.
6)The program also has autoland functions, and can calculate 'Phantom' VORs, ADFs, or using the GPS approaches.
7)All used navigational info would have to be confirmed manually for the required area to be flown, as MS is not renowned for accuracy of Nav info. This would also allow compliance with accuracy requirements for flight.
8)Legislation in Australia even allows flight within controlled airspace as long as certain criteria are met.
9)If all else fails, a failsafe system could be utilised, ie:Parachute which could be used if required for retrieval.
BIG project!
Any opinions and flaming accepted gratefully.
Rob.
Australia.

[wowo44]

So are you saying that all attitude/flight control calculations will be done with MSFS? If so, I don't think this will work as you are not likely going to be able to get the MSFS flight model to match exactly with the real aircraft. Additionally, you cannot simulate the actual conditions of the physical environment in MSFS. Over time, the difference in the actual vs. simulated attitude of the aircraft will diverge.

For example, what happens if the UAV gets hit by a gust of wind which causes the UAV to roll 60 degrees to the left? MSFS will have no way of knowing this unless you send the actual attitude information from the UAV to MSFS.

[rjfon1]

That's the way it would work.
If a gust hits it and sends it on a 60degree left bank, then this signal is returned to MSFS and then MSFS takes the necessary action being a right aileron.
The same goes for the attitude. The signal is sent from the real aircraft to the FS, for example a sudden gust causes a pitch up, this is returned to FS which takes the required action, being elevator down or power up.
This is the way computers work when they are networked in simulators. One does autopilot, one does radios, one does weather etc.
This is a very simplistic explanation, but alot of programming is required.
Rob.

[Unterhausen]

I seem to remember a group at Cornell University doing something like this. The thing is, you're increasing weight and power requirements quite a bit right off the bat. The job can easily be done in an ARM processor, and you've limited yourself to fairly high powered pentium class machines.

the advantage being that you probably can get money from Microsoft, which was my interpretation of the motivation Cornell had.

[vespa]

The challenge in designing an autopilot is in the incredibly complex methods needed to measure the attitude. Once the attitude is known, applying stability and navigation routines is so absolutely trivial that even the little anti-theft processor imbedded in your car's ignition key would be overkill. MSFS could surely be adapted as the ground station GUI for route planning and telemetry monitoring but you'd still need much more software on the ground to fully interface with the autopilot.

[rjfon1]

Yes I agree that the MSFS would be an overkill, but it doesn't have to be a high end CPU. This is only required due to the high requirement for a graphics angine. Without graphics, the CPU need is low. The groundd unit, however, should be fairly fast in order to produce the graphics display that shuld be seen. this can be displayed in conjunction with the returned camera vision.
Rob.

[treehog]

As a non computer expert joe soap learning the black art IMHO with MSFS I suspect but could be wrong that it will be difficult to work with compared to flight gear www.flightgear.com or YSflight www.ysflight.com which are open source code flight simulators

With microsoft you probably will not have access to the the source code and I dont know if API (aplication program interface )are available either so as to be able to interface the simulator

I go with the idea take a suitable flight simulator preferably one like YSflight which has a rc model jet simulator and a stripped down laptop

OK its overkill and will require a larger model to lift it but idiot proofs alll the other linkslike networks or sensors through USB and I suspect that logic will tend to succeed better

In the OS I suspect the MAC laptop g4 will be more stable and software crash proof but a faily modern laptop with a reinstalled XP (and service pack) with no other work to do should be fairly stable
not sure I would be soon keen on 98 and ME YIKES but 2000 might be ok also

Still difference is the spice of life so mayby the MSFS is the way to go but unless your a programmer guru and know your onions I would check with programmer experts can you interface with MSFS

If it does work I would be interested in the results to copy it but I suspect I would go with another flight sim such as previos listed above and put my MAC g4 up there

good job I got another laptop and soon getting the new MAC

Ralf

[mzk]

Hi well i am from india and an engineering sudent.I have been working on the same idea that is to use the simulator for controller.As i have worked on i believe what RALF says is in fact the best suggestion.
When working with MSFS you are limited to use windows as environment and to be frank when it comes to realtime and embedded systems windows is crap.You cannot make it realtime.Now if you choose linux environment the advantges are as follows:
1.you can use a real time kernel
2.you can go for any processor boards (small and compact) as log as ther are 32 bit atleast (check out pc104 boards they are real small and efficient too)
3.nextly you can use flight sim like ysflight or flightgear, now this will give freedom to tune the fs according to your model and then make it compact
4.when you work on linux i believ interfacing with any sensor would be easy.
5.this can reduce your codig time as you can incorporate pre built project codes (remember its open source so ultimate freedom)

order of the day i agree with ralf but use linux kernel

zaki