Automated Airfield Approach/Departure & Landing System

[FirmamentFX]

Ok - as promised in the mission programming language (http://www.rcgroups.com/forums/showthread.php?t=575890) thread, here is the approach/departure and landing system thread :D

As far as I can see, the basic requirements for the system as as follows:

- As little hardware as possible. ie using GPS for the position indicators for circuit flying (as long as the circuit is big enough, +/- 15 metres error in lat and long is not critical)

- Unless it is a permanant installation, the hardware (ie transmitters/beacons) that is used must be compact, simple to set up and take down, and ideally self powering (using a combination perhaps of battery and solar charging).

- The frequencies used must fall within the ISM unlicensed frequency bands (but have significant error correction and anti interference properties), and MUST NOT interfere with any commercial/military aviation setups.

The ideal is to have data for airfields stored in a database on the aircraft (see the other thread for details about this db - I am thinking MySQL) for the circuit loc, runway length and characteristics, and departure and approach routes.

Thoughts/ideas?

Cheers,

Martin

[FirmamentFX]

Forgot to say - the methods of achieving this as far as I can see are:

- Radio
- Ultrasound
- Microwave

I am tending towards microwave. I will post more details later.

[LukeZ]

How about this approach - instead of storing data on the aircraft, could you just make the system flexible? Set up a handful of beacons on the desired runway, 100 feet apart or something (length apart would set descent angle). The plane would know that when it reaches the first beacon it needs to be at 30 feet from the ground; the second, 15 ft; third is landing. I don't even see that GPS would be required.

I'm not a triangulation expert, so perhaps you'd need six beacons instead of three - one on each side of the "runway."

The hardware could be pretty simple - reads the beacon signals and controls the elevator and rudder servos. Have it be turned on or off with a spare R/C channel (or TTL input for integration with other autopilot devices). Simply line up the plane downwind, and hit the switch. The device would use the elevator to control descent, and rudder to maintain heading. An FMA CoPilot would be used to keep the wings level. It would calculate its location in 3D space based on the signals from the beacons.

Would something like that work?

And even if you wanted to go all out and involve MySQL in this task :rolleyes: , it still seems like this would be a good starting point.


Luke

[FirmamentFX]

Ok, here is the basic, simplified idea for lateral (centerline) guidance. The same principles can be applied for vertical (glideslope) guidance.

See attached jpgs for a visual explanation.

The transmitter at the departure end of the runway transmits a scan beam back and forth across the runway at a predetermined angle. The receiver in the aircraft picks this up as, essentially, a continuous signal along the time axis of the graph.

We know the frequency of the Tx scan (1 full cycle back and forth) as t.

When the time between the peaks of the signal (time = mu) is exactly t/2, the aircraft is aligned to the runway centerline. Equations can be applied to gauge the angle of incidence to the "horizontal" because we know t and the angle of the scan.

That, in horribly simplified form, is it.

M

[FirmamentFX]

How about this approach - instead of storing data on the aircraft, could you just make the system flexible? Set up a handful of beacons on the desired runway, 100 feet apart or something (length apart would set descent angle). The plane would know that when it reaches the first beacon it needs to be at 30 feet from the ground; the second, 15 ft; third is landing. I don't even see that GPS would be required.

I'm not a triangulation expert, so perhaps you'd need six beacons instead of three - one on each side of the "runway."

The hardware could be pretty simple - reads the beacon signals and controls the elevator and rudder servos. Have it be turned on or off with a spare R/C channel (or TTL input for integration with other autopilot devices). Simply line up the plane downwind, and hit the switch. The device would use the elevator to control descent, and rudder to maintain heading. An FMA CoPilot would be used to keep the wings level. It would calculate its location in 3D space based on the signals from the beacons.

Would something like that work?

And even if you wanted to go all out and involve MySQL in this task :rolleyes: , it still seems like this would be a good starting point.


Luke

You posted just before me! Will look at this.

M

[FirmamentFX]

Yes that could work. The only disadvantage I could see is the need for beacons to be spaced out before the runway, which is a factor when considering the "airfield" limits.

I wasn't going to use GPS for the landing - it's too innacurate - but to define where the circuit is so the aircraft could fly a circuit.

Martin

[GeoffH]

Ok, here is the basic, simplified idea for lateral (centerline) guidance. The same principles can be applied for vertical (glideslope) guidance.

See attached jpgs for a visual explanation.

The transmitter at the departure end of the runway transmits a scan beam back and forth across the runway at a predetermined angle. The receiver in the aircraft picks this up as, essentially, a continuous signal along the time axis of the graph.

We know the frequency of the Tx scan (1 full cycle back and forth) as t.

When the time between the peaks of the signal (time = mu) is exactly t/2, the aircraft is aligned to the runway centerline. Equations can be applied to gauge the angle of incidence to the "horizontal" because we know t and the angle of the scan.

That, in horribly simplified form, is it.

M

How would the aircraft decide if it was left or right of center line though? If a cycle = t center = 0.5t then the max course deviation on each side = t. That is if t was measured from peak signal to peak signal using no other means then the duration bettwen the signals peak.

How about a setup like a real VOR with 2 differnt transmissions 1 as a referance going off at known rate and the other scaning like you mentioned. Measure the time difference between the base signal tx and peak of the scaning tx to determine course. If peak rx is 0.5t after referance then on courese, 0.1t after referance then side of course, 0.9t after referance is other side of course

[FirmamentFX]

Good point! :-)

If 2 Tx's are going to be used though, then wouldn't it be easier to recreate a standard ILS system rather than a VOR, so there is no need to have a scanning signal? (see attached jpg).

2 Yagi aerials (each with a 40 deg arc or Tx), spaced apart 30 deg (each 15 deg away from centerline). Transmit on the same frequency, but modulate the signal by seperate frequencies. The receiver on the aircraft measures the difference in the depth of modulation od the 2 signals, and when the difference is 0 the aircraft is on the centreline - because the left and right signals are modulated by different frequencies, it would be easy to tell if the aircraft is left or right of the centerline.

[FirmamentFX]

Just found yet another way - using timed difference of arrival:

http://home.att.net/~jleggio/projects/rdf/tdoa2.htm

[eflightray]

What about lights as per aircraftcarrier landing lights.

Sorry, just passing and wondered what was going on these days.

[FirmamentFX]

Hey eflightray!

That would work for manual landings (assuming a video link between the UAV and the ground), but I'm looking at totally automated "hands off" landings (ie with autopilot settings: ILS APPR, LAND, FLARE, ROLLOUT etc).

One other thing to mention about the system - it should be degradable; that is, even a UAV without high-tech gizmos and computers on board should be able to fly the ILS approach with a basic telemetry system and just a couple of radio receivers.

Martin

[typicalaimster]

I've been tempted to setup some sort of VASI lighting system.

[Tuner]

Using something like the CMUCam2 and basic machine visoin in conjuction with special lighting system or retroreflectors could be used???

I think this may have already been done a few times. As fare as machine vision goes this is about as simple as it gets. Blob analysis find the centroid use several points to calculate the ground plane.

The key difficulty is illuminting or filtering the illuminated points in a way to clearly distinquish them from the background. Near infrared is often good, as low end cmos cameras can still be used.

One could concievable even fine tune a light low-power and low-tech setup that uses Thermopile arrays to achieve this. Using a narrow band of light off the visible spectrum along with more narrow field of view for each thermopile. If you modulate the light sorce you could us a PLL to enhance the signal from any background noise. You would need to use some creative engineering to get this to work reliably and be robust but after seeing how well 4-5 thermopiles work for leveling an aircraft outdoors I think it could be possible.

I would get 25 small thermopiles with say a 20-30deg viewing angle max. Arrange them in a square 5x5 array. The array wont be a flat plane more so resembling the surface of a TV tube or sphere where each thermopile is angled 10-15 degree from each other.

Have this array pointed forward and as the plane approches the lights it will stay pointed towards them. If the lighs are say 8 lights one at each corner and side of the field as the plane gets closer the array will begin to distinguish between each of the 8 lights especially if they have different modulations and the onboard logic will use these cues to control the plane much like the Leveler or horizon sensors.

GPS would be required in my opinion as well with both these techniques as they are just for the fine tuned control needed when the plane is close to the ground.

My 3cents.
Scott

[Unterhausen]

Not really on topic, but lotsa people have been working on vision for landing control:
http://citeseer.ist.psu.edu/653634.html
If you have a small enough airframe, vision seems like one of the few workable approaches at this point in technological development.

[FirmamentFX]

Using something like the CMUCam2 and basic machine visoin in conjuction with special lighting system or retroreflectors could be used???

I think this may have already been done a few times. As fare as machine vision goes this is about as simple as it gets. Blob analysis find the centroid use several points to calculate the ground plane.

The key difficulty is illuminting or filtering the illuminated points in a way to clearly distinquish them from the background. Near infrared is often good, as low end cmos cameras can still be used.

One could concievable even fine tune a light low-power and low-tech setup that uses Thermopile arrays to achieve this. Using a narrow band of light off the visible spectrum along with more narrow field of view for each thermopile. If you modulate the light sorce you could us a PLL to enhance the signal from any background noise. You would need to use some creative engineering to get this to work reliably and be robust but after seeing how well 4-5 thermopiles work for leveling an aircraft outdoors I think it could be possible.

I would get 25 small thermopiles with say a 20-30deg viewing angle max. Arrange them in a square 5x5 array. The array wont be a flat plane more so resembling the surface of a TV tube or sphere where each thermopile is angled 10-15 degree from each other.

Have this array pointed forward and as the plane approches the lights it will stay pointed towards them. If the lighs are say 8 lights one at each corner and side of the field as the plane gets closer the array will begin to distinguish between each of the 8 lights especially if they have different modulations and the onboard logic will use these cues to control the plane much like the Leveler or horizon sensors.

GPS would be required in my opinion as well with both these techniques as they are just for the fine tuned control needed when the plane is close to the ground.

My 3cents.
Scott

Now there's an interesting concept! I hadn't thought about machine vision. What kind of pickup range would there be for that?

I agree that GPS would be needed as well. I suppose the GPS could guide the aircraft into the glideslope/LOC pickup area and then the lighting system could take over.

M

[Tuner]

A more complex vision system could use the acutual image of the runway this is fare more complex but may work as well or better.

Exactly GPS to get it lined up with the runway then the Vision system takes over.

[TheBobby]

With new GPS, we can hope having a 2 meter horizontal position. With next generation of inertial systems (INS) for our size of UAV (an INS that doesn't exed 50g), that mix GPS data and other sensor with the right kalman filter, we can hope having a 20cm horizontzal precision. If this is possible, we can use it with a precision altitude sensor system (ultrasonic, or other, i dont know).

The combination of those systems can provide good results but will be expensive.

That's just a point of view, mine ;)

[Accu157]

Tuner, are you saying you'd use a filtered section of light? I'm thinking there might be some telescope filters which could narrow down to the spectrum you want. Cool way to triangulate direction and location.
How about something like how a loran station works? Would the refresh rate by high enough? I'm not a UAV guy, but the science and tech is really interesting. :D Perhaps you could use high-frequency auditory cues.
http://en.wikipedia.org/wiki/LORAN

[Tuner]

I worked on the conceptual design of a similar vision system. You would use what is called a band pass filter or a shelving filter in some cases.
Edmund Scinetific is a great source of these filters.

Modulation is an excellent way of removing background noise but wont work well with video.

Certain Wavelength are not very abundant in nature and can be used Infra red is the most commonly used. With the right filter combination you can work wounders but filters cause problems as even the best filters filter out a good percentage of the light you are wanting to see and so using very bright sources is KEY!!!!!!!!!

Strobes would be a great intense source the strobe pulse would have to last for 11/2 times the length of a video frame if you use video so that you have one frame fully exposed to the strobe. If you use an omni light source youll need some juice.

I keep waiting for these new cameras to come out that should make tasks like this much simpler lower power and smaller.... www.anafocus.com