blimp autopilot indoor

[SOOline]

New to posting, just wanted to say thanks to all the contributors who make rcgroups such a valuable resource. My question, I'm new to the UAV thing and wanted some seasoned opinions on the best way to execute rc blimp autopiloting (point to point navigation) indoors. It would seem like there should be a way to merge some of the systems i've been looking at that use GPS with some of this indoor GPS technology I've been readin about. It would also need to be scalable as I would like to have many blimps in the air simultaneously.
Thanks in advance.

[hugo_vincent]

Indoor GPS is very much dependent on the building/room you are operating in. Before you pursue it's use further, I would strongly recommend buying a receiver that does indoor GPS and testing it extensively in the room/building where you plan to fly your blimp. Indoor aerial navigation seems to be a pretty difficult problem (compared to outdoor - you have to worry about walls/floor/roof, lack of good GPS (even if you get a good fix, the accuracy of many receivers may limit it's use), etc.). Computer vision might be a good approach, but of course that is difficult too (to say the least)...

Good luck!

Hugo Vincent.

[powerblimp]

i will say the same
indoor GPS does not work and accuracy is not enough
work on CCD cameras and IR or pietzo systems

good luck !

regards minizepp.com

[Tuner]

Their are Motion tracking systems that might more suite you needs.
RF based Tracking sensors have a long range some up to a KM.

Also the North Star is an "Indoor GPS" that projects a pattern on the ceiling and all robots use a fisheye lense on the top and they calculate their position based off of known aspects of the projected geometry.
A cheap version of the Northstar would work but be difficult to create the software for or expensive.

[Ailuropoda]

Off hand, I'd be tempted to try one of three things.
The first one would be to simply stick up beacons where I want the blimp to go, which may or may not be practical given the type of indoor flying you want to do.
Next thing I'd probably want to try is building a hyperbolic positioning system using ultrasonic.
If those two didn't work, the last thing I'd do, which really would be the last thing I'd want to try, is to put cameras on the ground pointed upwards and compute the position from a ground station.

[Fabian]

You may built a mini-radar to avoild collision.

[SOOline]

Wow, thanks for all the great suggestions.
I've been doing some more research and it doesn't look like any sort of actual indoor GPS is going to be realistic. I've thought about the beacon idea, some how plaving markers on the floor and having the blimps flying from marker to marker. I've researching RFID technology but it is still unclear weather or not that would work over the range that I'm considering (about 150 ft by 300 ft.)

[SOOline]

Tuner do you have a link for the North Star project you were mentioning?

[Tuner]

Yah and hear are some Trackers.

http://www.evolution.com/products/northstar/

http://www.polhemus.com/ Ooops To small of a coverage area.

I bet you could find a system like the Polyhemus with sat +- 1inch accuracy or even less that would be very suitable.
I would call Polyhemus chances are if they can not expand the range of one of their systems with loss to accuracy then they can tell you of somone who can. They key is you only need say +- a few feet accuracy.

Good Luck Their are other product out theri like the Star Track RF wireless motion tracker. Some as I said have a very large range they are used in Indy Races and such(noth cheap).

I have considered Buying a large wharehous and filing it with obstacltes. Then build a bunch of blimps shaped like submarines and fit them with wireless video and ultrasonic sensors for combat. It would be like a video game only real. Fill em with hydrogen and let them blow up, that would be fun! Ok gota go....

[ElectroLawndart]

Another possiblity is to use a ground based, stereoscopic vision system ala Robo Soccer. Use the processing power of a computer on the ground to track your blimp in a 3D space and uplink steering commands. Maybe instruct the computer to steer the blimp in a circle centered in it's field of view. At least that way you don't have to worry about blimp knowing where it is.

Dart

[Joe2003]

You could use a stamp processor and an array of rf range sensors? I have seen people creating a full 3d image of a room passed on this setup. Also the whole thing would weight in at only a few grams without battery.

[SOOline]

http://www.ubisense.net/

This may be the best bet I've found yet. Just wrote an email to the company concerning pricing. Tell me what you think?

[lucent]

if it is just an air ship your talking about i dont see why you couldnt program a vector and time to execute a turn. it should be possible with no external hardware. only an internal compass.

lots of pillars and you'll be out of luck.. you could of course just have a computer record stick movement and play it back. so you make 1 static run and let it repeat. this probably won't work with a fast mover, but an airship.. it should be ok..

[SOOline]

So the ubisense link I put up looked really promising but 12,000 dollars for the academic starter kit it is well out of my budget range.

Lucent I like your suggestion, however in my mind it really seemed like the inertia and other chaotic variables would cause the blimp to move significantly off the desired path after a period of time. Again this is a 30 to 45 minute show so the integrity of the ppath would need to be maintained over that period of time.

I recently came across this
http://www.ias.uwe.ac.uk/People%20Pages/j-welsby/main.htm
and these researchers have managed to get the blimps to track an infrared beacon. Now if I could set up a grid of infrared beacons of different wavelengths then I could possibly turn them on and off dynamically throughout the performance. Any feedback?

Also I'm blogging about my research for this project, if anyone ants to check that out it is

http://project-condor.blogspot.com/

Thanks so far for all the great suggestions and feedback.

[kd7ost]

if it is just an air ship your talking about i dont see why you couldnt program a vector and time to execute a turn. it should be possible with no external hardware. only an internal compass.

lots of pillars and you'll be out of luck.. you could of course just have a computer record stick movement and play it back. so you make 1 static run and let it repeat. this probably won't work with a fast mover, but an airship.. it should be ok..

This might be a little dicey to pull off. I fly an RC advertising blimp at the BSU (Boise State University, Yep the Smurf Turf College) Arena for the Basketball season. You wouldn't know it by sitting there as a spectator, but once you put a neutrally balanced lightweight floating device in the air, you find wind currents. Depending on heat, air conditioning, circulating fans and even convection currents that thing won't sit still. It will rise, sink and turn etc all on its own. You might have to have a fairly active control system, or keep it to a small room with pretty dead air movement. Think of the air as a thin fluid, and it is in motion.

You also have to stay pretty anticipatory with the thing. It's a little weird at first. I liken it more to operating a RC submarine in very clear water. It won't turn on a dime or change altitude, direction whatever. It moves slowly so you have to be patient and let maneuvers progress. If you make a 90 degree turn, you have to be out of the rudder or go counter rudder pretty hard before the turn is even complete. Momentum keeps it on it's last course for a while. It has a lot of very subtle pilot inputs all the time to keep it out of lights, poles, stands etc. I even learned the currents in the building so I know when it will start to sink, rise, turn etc. You might have to deal with issues like that in the control system.

Dan

[Tuner]

The infrared beacon is great if it is completely indoors. You can use infrared frequencies that are not emitted by the lights. OR Ultra Violet but uv sensors and LED can be expensive.
You could use an octahedron or dodecahedron array or 8-12 photo detectors. This array can allow you to locate beacon in 3d space you may need more it depends on the sensors field of view.

Modulate each light source so if you want blimp "A" to go to point 5 have point 5 blink at Xhz in correspondence to the blimp "A" modulation.

This is really easy because you can use the whole Moth technique where a moth’s wing opposite of the light source flaps faster causing the moth to fly into the flame.

OR
You could have say 3-5 light sources all modulated the same or not at all
Then you use array of sensors to triangulate the position of the blimp. It could be rather inaccurate unless you increase the sensors in the array or use a camera but it seems that is not a big deal, but you never know.
Me personally I would use two cheap CCD IR cameras with filter. Modulate the light source to distinguish it from interference.
I would put fisheye lenses on both and place them back to back. Then lay out a 3-5 light sources and program in their positions. You can use some free Machine Vision software that with the known position of the points triangulate its position.

This is basically a homemade Northstar. Much less accurate but you don’t mind.

I can point you to LED's Filters Cameras Lenses and Sofware if you are interested. I have done some similar work in tracking IR light sources.

Scott

[SOOline]

Scott I would definitely like any information you can offer about the LEDs Filters Lenses and Software. I think I'm following what you are describing, Do you think your method would be scalable? Eventually I want to be able to track and control 10 to 16 blimps simultaneously.

Also if you have any publications or documentation on the IR tracking I would like to read it.

Thanks
Casey

[SOOline]

Dan thanks for the first hand information about flying indoors. That confirms what I thought about the controlablilty of the blimps indoors. How big is the blimp you are flying? What kind of lift does it generate?

[kd7ost]

Dan thanks for the first hand information about flying indoors. That confirms what I thought about the controlablilty of the blimps indoors. How big is the blimp you are flying? What kind of lift does it generate?

It's a 10 foot Standard RC indoor blimp from these guys. http://www.southernballoonworks.com/rcblimps.html#STANDARD_INDOOR_ELECTRIC_RC_BLIMPS

Lift is perhaps marginal. That's a tough question for me since I never did check it. The lift can be varied though. The vinyl stretches. You can put more cubic feet of Helium to get more lift. The lift ratio to helium fill is exponential. Once it overcomes the weight of the blimp vinyl and gondola, it doesn't take as much helium to achieve higher lift levels.

I have to get it topped off and neutral balanced before all the games. I have to top it off as the helium molecules do escape through the pores in the vinyl. I overfill it while it's tied down with a lanyard.

The nose and tail of these units have a small vinyl bag velcro'd in place. I add pennies to these bags until I get the blimp to sit neutral. It usually takes a little flying around even after that to get in different air to final tweak it. I might add one penny or take one out of the nose or tail to get it the way I want. It takes a lot less power to fly around if you don’t have to deal with it trying to rise or sink on you. It's more responsive too.

I use as many as 60 pennies or more to get the balance right. (How much does that weigh?) They also supplied NiMh battery packs of 5 cells. Those could be replaced with LiPo's to get more lift. Bigger blimps take more helium and have more vinyl. I don't know how much lift you could get going bigger. I do know I could have put in more helium and stretched the vinyl more. If you leave off the graphics, that would also save weight. It's just extra vinyl bonded to the existing envelope. They also supply it outfitted with two standard sized RC servos. The fan direction servo can be switched with a slow speed high torque micro servo. The full size servo used to drop coupons, tickets etc can be left off entirely. They also use a standard Futaba 72 MHz receiver. A micro receiver would help.

I also see they have new upgraded units with brushless motors. That would give you more response. I do know that the lighter I fly it at, (less pennies) the faster it responds to the marginal power system. If I have two games in a week that are 3 to 4 days apart, I fill it really full and put in a lot of ballast before the first game. Then later in the week I only have to remove pennies to compensate for the slight helium leaking. It is more responsive when the ballast and resultant inertia is minimized. After a week and a half I do have to top it back off. It starts to wrinkle as the material sags. BSU lets us keep it inflated and stored in the upper level.

Those newer brushless motor units they have are probably a lot more powerful. That would help you out when you don't want to spend 30 seconds of painfully slow time to make a turn and keep at altitude while flying.

Your system will have to be pretty exponential, with a fair amount of hysteresis in its response.

The more I think about it, the easier I think it is to have an RC airplane Fly at 60 mph via GPS waypoints that what you are working towards. I think the project sounds very intensive. It will take a lot of work. I’m very interested in this though. It’s tough robotics.

Dan

[kd7ost]

I should say that with a generous amount of aggressive flying, you can get to to move around pretty quickly. Turns are faster at full speed. I can crank a turn in while stationary, (tail fan causes yaw changes or spin to occur) put full speed ahead with the main drive fans pointed downward or upward and spin while descending or ascending. I can go from the catwalk in the rafters down to the lower seats in about 15 to 20 seconds while completing about 15 spins. That's about as radical as you can get it to move but I want to be careful to state it's not as lethargic as I may have described, at all times. You can spin it pretty fast. It is slower to get in forward motion or stop from forward motion.

Dan

[Tuner]

OK You number one resource online will be:
http://www.cs.cmu.edu/~cil/vision.html
www.machinevisiononline.org/

If i were to use a camera right now for something like this I would use the CMUCam2:
http://www.cs.cmu.edu/~cmucam2/

You will need a fisheye lense and technically 2 cameras back to back facing front and back for full coverage.
That said a guy could get it to work with one camera with a fishey pointing down this depends how high the blimps are but may work very well.
Lense and filters get at:
http://www.edmundoptics.com/
Great resource I bought a lot of stuf used it for a day and returned it because it would not achieve desired results.
For LED'S
http://www.roithner-laser.com/

If you want somone to do the work for you I recommend this company:
http://www.vision-comp.de/
Ask for Mr. Endre Toth I think I remember his name right he his the US rep.
Very helpful person.

If you want more specifics pease ask.

In general you want a system that does blob analysis that can give you the center of the light source and other info such as overa brightness, area, size and how many frames it is visible for, how far it moves over x frames......

Id love to know more about why 10-16 blimps would be in a small stadium though I can kinda under stand.

One issue that is why I think some sort of UV light or Modulation is require. "Flashes" from cameras could interupt things as I know they put off a good chunk of IR not so sure about UV

I like to stay around the 400-415nm wavelength in UV and 765-790nm for IR IF you go any lower or Higher you then run into having to use expensive camera CCD's that can detect UV and deeper IR. Though going lower or higher gets you out of the noise of everyday light. If you can pull off using Thermopiles/Photodetectors these invisible wavelengths are more feasible but the design would have to be much more cleaver to use an array of detectors and goes beyond me some as it would require more of a hardware approach but done with the proper person is the most economical minus the money youd pay for development. (all this is IMHO at least.)

Scott