poor mans autopilot

[pistaandrea]

Poor mans autopilot
Not enough money? You want to stabilize any type of airplane? Use what it is available on the RC market.
We made a really cheap set up to guide safely an RC model on a mission. Right. It is based on copilot – infrared system. We all know what it means. But look at the video taken from my real plane when I followed my acrobatic unstable model plane. The RC model behaved like a super stable jumbo jet liner on a route. Amazing.
We are waiting for comments, deciding whether to offer our product to the market. Go ahead.
The autopilot controls: speed, altitude, GPS position, altitude in relation with GPS position, attitude.
Standard cheap GPS receiver works as a navigation computer for the model plane. This means that it is able to fly only A to B routes, or A to B and back to A routes (with special procedure). When it gets to the end, it starts holding. The system can do action at specified location – trigger a camera, drop a load and so on.
Let me know, what you think.
The video is to be found here:
http://www.youtube.com/watch?v=uA6w5ow89Go

[koiravahti]

Is it possible to use this system for the Return To Launch system? What's the size and weight? How cheap is the price? Are you planning to offer it in Europe?

[pistaandrea]

Is it possible to use this system for the Return To Launch system? What's the size and weight? How cheap is the price? Are you planning to offer it in Europe?

Yes it is
The size is like two boxes of standard receiver (same weight) plus GPS receiver.
Price 700-1000 dollars, but no idea at the moment. Easy star had no problem with its weight.
Give me some time I am working on a website dedicated to the Autopilot system. There you will see

[Connexxion]

This is interesting.

Do you have some pictures of the autopilot?

I would also like to see a planned versus recorded flight path.

You said you're using a cheap GPS receiver that does the navigation.Do you mean a handheld one?Because these can handle more then a A-->B-->A route.

What size/weight was the plane?

Looks like you're in Europe wich could be a big advantage for you since the USA based autopilot manufacturers are limited in export by new rules regarding autopilot export.

I'm anxious! :D

[maguro]

I also would like to see photos.

[icebear]

Interesting!

I am not trying to be negative, but for the same amount "700-1000 dollars" you can get a Picopilot incl GPS with more features than you are describing.

Still, would be nice to see some more details of course...

/Icebear

[VasMan]

Are you planning to offer it in Europe?After czeching out the video, I'd say it's already there... :D

[zlite]

Yes it is
The size is like two boxes of standard receiver (same weight) plus GPS receiver.
Price 700-1000 dollars, but no idea at the moment.

Yikes. That's not a poor man's autopilot ;-) Just to give you a sense of the pricing bottom of the market, we're planning to release ArduPilot (http://diydrones.com/profiles/blog/show?id=705844%3ABlogPost%3A41212)in Sept for $29. That's without a GPS module and you have to add your own FMA co-pilot, but when you include them both it's still less than $150.

Later in the year, our ArduPilot Pro (http://diydrones.com/profiles/blog/show?id=705844:BlogPost:40459) with built-in GPS and thermopile stabilization will probably sell for around $100.

All open source code, including ground station.

It looks like you've got some terrific technology, but like Icebear I'd encourage you to look at the pricing again.

[pistaandrea]

Well thanks for your comment. I am using copilot from FMA as well as you. Regarding the expected price – I am not sure about it, but once I is decided, it needs to be respected. The potential buyer decides himself if it is worth his money. My friend and I were working on it just for fun. We both have well paid jobs. There is no need for us to be anxious about selling our product. If somebody likes the way our autopilot handles un unstable acrobatic plane, he might want to have it even for such price.
It is great you can make your own system so cheep, but it is not available yet is it? Has it been tested? Did you follow a similar model plane as mine in an airplane to see what it does on a route. When you are ready to offer your product for such money, I will be the first to get it. Regarding pico-pilot – I had it and returned it to the producer. It had no chance to control plane like that one I filmed.

[zlite]

I do understand your position, but was just curious where your price estimate came from. The electronics in an autopilot (CPU and other components) typically cost less than $20 and the GPS maybe another $50. Add a few more dollars for assembly and a box and maybe you can get to $90....

But how do people get to prices of $1,000 or more? Is it just charging for intellectual property (software)? I come from the open source world and that's all a bit foreign to me. We charge for hardware ("atoms"), including a margin for our manufacturing and retail partners, but give away software ("bits").

Our autopilots will available commercially by Sept. But no, we haven't tested them by flying a manned airplane next to them, nor have we flown them with acrobatic airframes! That's not really the intended use mode ;-) We're aimed more at hobbyists who just want to get started with something cheap, easy, and hackable.

[hoysome]

cheaper and more available way to "follow" a plane in testing is with camera gear onboard. either a recording device, or a transmitter onboard to watch and react live.

[CenTexFlyer]

Probably the biggest drawback to this type of system is the inherent limitations of a thermopile driven stabilization device. It doesn't work over snow, water, sand, or on an overcast day. If you are spending a sunny day at the park, then that's OK, otherwise........

[zlite]

Probably the biggest drawback to this type of system is the inherent limitations of a thermopile driven stabilization device. It doesn't work over snow, water, sand, or on an overcast day. If you are spending a sunny day at the park, then that's OK, otherwise........

Are you sure? We've had no trouble on overcast days and over water (haven't tried sand or snow). Given that thermopiles are the basis of Papariazzi and AttoPilot, both of which are competition-grade autopilots, I feel we're in pretty good company.

Please note that we use onboard accelerometers to calibrate our thermopiles at startup, so this is better than just using the FMA Co-Pilot.

Do you have experience with them yourself? I'd be interested if someone had data on this issue, since we don't see evidence of the alleged problems ourselves.

[Kaptah]

A decent autopilot is surprisingly expensive to manufature. Below are approximate prices.

- CPU: 10$
- sensors (acceleration, gyros, pressure): 100$
- PCB: 8$
- GPS: 40$
- Other components: 15$
- Automatic SMD assembly in a factory (50pcs, machine programming work + assembly work = 2000$): 40$ per PCB

TOTAL = 213$

+ Solder paste stencil for one side SMD assembly (can be used for many times) = 200$.

[zlite]

A decent autopilot is surprisingly expensive to manufature. Below are approximate prices....

Well, maybe you don't consider a thermopile autopilot "decent" (tell that to the Paparazzi folks!) but here are our costs:

--CPU (2x 20 MIPS Atmel (http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=ATMEGA168-20AU-ND)): $4
--6 Thermopiles (@$4.77 each (http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=MLX90247-ESF-DSA-ND)): $28
--PCB: $8 (same as you, although our price is actually $2 at volume, and I'm including two small daughterboards for the thermopiles)
--GPS: $40 (same as you)
- Other components: $12 (about the same as you)
- Automatic SMD assembly in a factory: $6

Total: $98

The big difference is that our sensors are cheaper than IMU components and our SMD assembly is a lot cheaper. Granted, our boards are very simple and we supply all the non-SMD parts (just connectors) unsoldered, so the user has to attach them themselves, but I think if you work with a bigger manufacturer you can get a much better assembly quote.

[Kaptah]

I have nothing against thermopiles, I just don't know them. Very interesting technique, that is. I've been wondering, what is the short term angle accuracy (compared to the horizontal plane), you can get out of using thermopiles? Do you have to use averaging?

Are you using one AVR for servo control and the other for other stuff?

Yes, I agree that the assembly cost here is outrageously high (actually you could almost multiply those $ values by 1.6 Euros...). This is mainly due to labour related cost in the factory. The cost gets very high, if you don't deliver your components in reels. Then they assemble the components using pincers...

[dmgoedde]

"Regarding the expected price – I am not sure about it, but once I is decided, it needs to be respected. "I must say that the market will decide for you! Something is worth what people are willing to pay for it. Charge too much and the sales volume is choked down, and total revenue and profits are small. Charge too little and the sales will saturate with minimal revenue also. There is a sweet spot in price.

[zlite]

Charge too little and the sales will saturate with minimal revenue also. There is a sweet spot in price.

Absolutely true, if you're a profit-making business. But this is a non-profit for me---I just want the maximum number of people to be able to try and experiment with autopilots (hopefully both expanding the commercial market for higher-end products such as yours, and creating a larger developer community that will make our own open source projects easier).

In my world, cheaper is better: we're not looking for the profit-maximizing price--we're looking for the market-maximizing price.

[zlite]

Are you using one AVR for servo control and the other for other stuff?

Yes, one Atmega168 is for inner-loop real-time stuff (mostly stabilization with the thermopiles) and the other is for outer loop navigation and telemetry stuff. We could probably do it all with a single ATmega but all the extra processing capacity of the second processor allows people to add features and functions without disturbing the really important innner loop timing.

BTW, there is actually a third AVR processor onboard, an ATtiny45, which just controls the failsafe. We could have done that in software, but wanted it to be a stand-alone circuit for safety reasons so that an autopilot crash wouldn't affect our ability to regain manual control.

It's about 50 MIPS total onboard processing power, which isn't bad at all for $5!

[dmgoedde]

Absolutely true, if you're a profit-making business. But this is a non-profit for me---I just want the maximum number of people to be able to try and experiment with autopilots (hopefully both expanding the commercial market for higher-end products such as yours, and creating a larger developer community that will make our own open source projects easier).

In my world, cheaper is better: we're not looking for the profit-maximizing price--we're looking for the market-maximizing price.Chris - my price comment was aimed at the thread originator declaring a $700-$1000 price and Bjorn's great reply that for same $ someone could buy a PicoPilot which has more features and GPS included.

I edited out my GPS price comment, but especially since your goal is to get the Ardu Pilot into as many hands as possible, I have a source of GPS that is about half the price you mentioned, and this company offers several chipset versions of GPS (MediaTek, Atmel (UBlox), and SirfStarIII) and their quality and customer service is outstanding. This could help you shave another $20 from your price and get your system into more hands.

[zlite]

The cost gets very high, if you don't deliver your components in reels. Then they assemble the components using pincers...

Yes, which is why we deliver all our SMD components in reels ;-)

[pistaandrea]

Some of you requested more details on our system. I created a fast web with some pictures.
Have a look.
www.volny.cz/leteckafotografie/Autopilot

Regarding the price discussion, I thing it is not worth to discuss it until the capabilities of an autopilot has been proved. The most expensive think is the software development and the testing. Our autopilot flies and was tested on different airframes including flying wings. The most difficult problem to solve for us was altitude. There was always oscillation. Finally there is none. It can deal with surprising turbulence now.
Thermopiles can deal with much more weather changes than advertised.
Our plane was flown on tracks longer than 100 miles, starting early in the morning and arriving 2 hours later. During that time the temperature changed considerably. There was no problem as well as over big lakes and snow. I believe it would not work in clouds or fog. Anybody likes to fly in them?

[Connexxion]

Hi Pistaandrea,

I was wondering if you tested the other autopilots you had (RCAP/PICOPILOT/PDC-10) also with the FMA Co-Pilot for levellingpurpose?

Furthermore,how do you control altitude?
By throttle or elevator?

Does the autopilot hold it's altitude it was switched on or can it climb to preprogrammed altitudes?

Do you have a record of the altitudecontrol during flight?
(the garmin geko 301 has the ability to record the altitude during a trip by means of a built-in barometric altimeter)

Thanks in advance.

[pistaandrea]

Yes I did
PDC 10 was fine, but did not keep the altitude as our AP.
RCAP had several issues that I reported before. Main thing was not enough power for the rudder. Big planes were in trouble. The wind blew the rudder in to 0 position. No steering, we lost several planes. With small ones it worked.
Picopilot was useless for my purposes. I hated the way point editor.
The altitude is programmed and the plane limbs there alone.
Altitude is controlled by elevator. Yes a have a record of the altitude, it is even with no oscillation.

[dmgoedde]

Yes a have a record of the altitude, it is even with no oscillation.I think Connexxion was asking for you to post the altitude data here for him to see.