Micropilot reliability

[strixi]

Hello!

I would like to know opinions on the Micropilot MP2028g reliability and configurability, since I have been working in the integration of this AP and have found many many problems.

First, the integration process is quite simple for a simple RC model trainer, but gets quite hard for planes with higher complexity in it configuration, like inverted V-tails, separate flaps, etc.

Second, I am seriously worried about reliability, since I have experienced several negative behaviors that after a full review of the system I concluded is out of my reach to solve, like unspected battery voltage decrease (believe me, without reason because batteries were full of charge and healthy before and after the incident), just to mention one. With this background, I don't want to even think about autonomous takeoff and landing. So please, I want to hear user opinions of the system (good and bad, both) in general. I would appreciate them all and I think it will healthy for all of us (users) to know how good or bad a system is working independently of what marketing stuff say.

Thanks very much

[djklein21]

I have found the system to be anything but reliable

[danstrider]

I won't say I've found it 100% reliable before I had it set up properly. I too experienced voltage levels drifting around; check for ground loops or look into common grounding your batteries. Also, you have to connect a wire to ground to measure battery voltages I believe, discussed in the manual somewhere. I experienced more wireless link issues than anything else. It was always a hassle to get the wireless link setup correctly the first time. After it's working, play with the time between packets, reducing it incrmentally until the transmissions go very poor and you know your limit. It's not an intuitive piece of hardware, but with three month's persistance, you can get it set up. Once it's working in the aircraft, if you have a trainer-style plane, the tuning process is painless. Are you competing in a student competition with the unit or is this industry related?

I too have been scared of autonomous t/o and landing with the Micropilot, but I have also seen it done quite successfully. Small steps and persistance is the key. Don't know what else to tell ya...

Dan

[Bg~]

I've heard from 3 different AUVSI University competition groups that used the micropilot and had a LOT of problems. However, I do know there is a guy at NASA who has used it to do over 200 autonomous landings (I'm a little fuzzy on the number, but I think it's at least this many). So...seems like it can be hit or miss..

[strixi]

My project is industry related, but we have done our first steps into autonomous flight with an RC trainer, just to learn in a safe way. What I am finding interesting about Micropilot is that people who use it either just manage to get it working more or less fine or they get stuck trying. I would say that when you buy an MP2028 you run a serious risk of not being able to setting it up properly in your plane configuration. I guess, from the opinions I have read, that this risk is almost zero with a system like Piccolo, which seems to be more "high end" stuff.
The point is that every Micropilot succesful story is usually associated with a 4-channel trainer, in which the tuning process is, as "danstrider" says, painless. In my experience, I agree with that because I haven't had much problems with my trainer, but things get difficult when you add complex things, like V-tails, engines with chokes, separate flaps, etc. Have anyone used a MP2028 in a complex plane (I mean, more than a 4-ch trainer) with success?
Besides, have anyone used it in long range flights? Could you tell me your experiences?
Thanks

[djklein21]

I would really enjoy hearing more about this topic

Thanks for all of your input

[danstrider]

Long range flights meaning out away from yourself? I've flown via a Micropilot for 35 minutes straight at 35-45mph ... you can figure out the distance travelled. It was within sight the whole time doing circles and various circuits, but it was really boring for the safety pilots :-)

I've never tried to put it in anything other than a 4-channel trainer style aircraft (dual aileron servos though) and others I've heard have also had problems setting up exotic aircraft configurations. It seems easy to set up flaperons, flaps, and other wing types through the wizard (I got flaps to work this way once), but I've never flown in other than just dual channel ailerons.

I'd love to try a Piccolo, but I haven't persued the educational deal after I sent in an application and didn't hear back from them (my fault, not a critique of Cloud Cap). I have seen the Piccolo do autonomous takeoffs and landings numerous times. A company out west (http://www.arcturus-uav.com/t15.html) has had very good success auto landing with the Piccolo.

Some of the AUVSI Student teams had problems layered on top of the Micropilot such as needing a seperate kill mechanism that the Micropilot does not provide stock. I have been in their shoes and I highly recommend a person dedicated to Micropilot installation and development. I also recommend reading and highlighting all the manuals (including the ones from the support site). It's a difficult system to get into when you have other things to worry about.

Also, I've found the payload control functionallity of the MP2028 to be both useful and limited. Sometimes it's really easy to set up what you want to do onboard; sometimes programming a thread like you want is pretty much impossible. I recommend seperate payload control from the MP.

Dan

[strixi]

Yes, with long range flights I mean far away from me and any safety pilot. I have also flown circuits above me autonomously for more than 20 minutes. But the question was oriented to flights in which you must rely on your autopilot and safety pilots have nothing to do. After all, that is the objective of any UAV, if not we would be using just RC models. Anyway, safety pilots don't like the idea, jaja! Are any long range flyers out there?

I agree with you, I would love to try a Piccolo autopilot. I have read it is much more reliable and robust plus easier to use. Also the Kestrel seems to be quite good, but I haven't read users opinions yet.

I have also tried the Micropilot with a 4-ch trainer and it was a painless job until you want to move further and do some complex tasks, like payload integration or advanced waypoint navigation. Unlike Piccolo, which has 2 serial ports or Kestrel that has 4 of them, Micropilot does not has any and only supports servo-type payloads and ADC inputs. And althought a serial port can be implemented as a general external payload interface, with the limited Xtender documentation it's a painful task and if you want people from Micropilot to do it, you'll have to pay extra. Have anyone done something like this with a Micropilot?

Esteban

[jescardin]

Hi friends:

My experience is from a professional point of view as I used to work during 10 years for a Spanish firm devoted to the design, operation and development of UAVs for the Spanish MoD.

The last five years we were working with Micropilot, first with its MP2000 unit and just from last year with the MP2028.

I may advance YES, it is quite difficult to make it work properly but we did for four different models (two turbine powered delta platforms and one propeller driven V-tail target aircraft plus one T-tail observation aircraft) although on the learning curve fell "important" persons like one Aeronautical Engineer, finally fitting the system between an Electronic and Computer Engineer and myself, then the company test pilot.

The turbine powered delta platforms aircraft (one was a twin turbine powered aircraft) were configured for engine, elevator and aileron MP control, while the human pilot retained always control over recovery chute and signalling smoke and IR flares.

The propeller driven V-tail was also configured as above, without rudder control.

Finally, the T-tail observation aircraft was configured with full control: MP controlled on automatic mode throttle, rudder, aileron and elevator. Beyond these controls, anything more you give to MP you loose direct control over that; this is, for example, in the observation aircraft we decided to keep flaps for human control for take off and landing rather than assigning this to MP as if so, the human pilot lose any control over the flaps (I know the same apply to the MP2028).

Regarding MP2000 operation you must know we recorded sea missions (aircraft launched from a Spanish Navy large ship) with the single turbine target aircraft with first waypoint well of over 50 km (30 miles) from the launch site and with a total range of nearly 100 km (60 miles) at speeds of nearly 300 km/h (180 miles/h) and at 300 m (900 ft) height all with acceptable success. Normal missions with V-tail propeller driven target aircraft are of about 50 minutes flying a “8” pattern at between 3 to 8 km (2 to 5 miles) from the launch site (I personally have seen militaries on charge playing cards while the aircraft was flying totally safe and only working on the computer when requested to change speed, altitude or recovering the aircraft.

Regarding autonomous take offs and landings, I am sorry the aircraft I produced were not intended for it but we "suffered" one when trying to take off with a reduced scale model of the T-tail observation aircraft and we suffered a control lost, and, as we had PCM programmed with MP2000 "ON", it sensed "0" height and not enough speed and applied full throttle till getting rotation speed when it took of perfectly!....and it was without the special take off/landing sensor!.

After reaching the correct altitude it started the programmed test "8" pattern. As it was quite late with poor light and everything was really an emergency, I tried to switch to manual again but every trying ended with the aircraft inverted!... so I had to go to automatic again and hope the MP2000 recovered the aircraft as it really did!. Finally when the aircraft passed over my head, I switched to manual and with a tight turn was able of landing the aircraft undamaged.

Regarding the false battery readings we also suffered them but were able to trace them to the RX and Servos volt meters misscalibration after uploading the MP2000 firmware. It also may occur after you load your model *.VRS archive as it contains calibration parameters, so it is necessary to make "000" so the micropilot recover the original factory calibration.

Last year the firm bought the first MP2028s against my opinion as it had taken a lot of work and effort to make the MP2000 works on the four models (all within a year and a half!) and we were still discovering new things and improving reliability at high speeds, but I was told Micropilot had ceased MP2000 production and had no units left.

Due to my critics I was put aside of the development and I was asked to work getting quality and military certifications for the aircraft. My workmates then on did an excellent job integrating on a single waterproof metal box the receiver, the MP20208 and other electronics but, unfortunately, were not much successful as end of year operations proved very troublesome. In fact every MP2028 equipped aircraft lost airspeed data soon after launching so making operation impossible or even leading to the crash of the aircraft.

The firm owner asked a solution to Micropilot, but as I was fired from the firm this January do not know if it was finally solved or not.
You may be sure that Spanish firm is one of the most experienced ones on the Micropilot product tuning up and operation even I well know of a very (very) large European aerospace company which gave up with MP2000 even counting with direct Micropilot support for their range of target aircraft.

From my personal point of view (I really got around 150-200 MP2000 flights and teached operating units within the Spanish Air Force and Army), major drawback of MP2000 was it was not totally able of aircraft trimming, so even we devoted hundred of hours about the trouble with comment from Canada, I got the idea the MP products simply are not conceived with full autotrim capability and it is most important to bear in mind when you think about production due to normal fabrication tolerances, and for a perfect service to your customers you should test fly every single airframe (even without autopilot) so you may record the trim and include on the servo "0" position.
That matter was what most troubles gave us as some aircraft, without an apparent reason, went circling without entering the programmed pattern and only if detected early on the flight they were able of being manually recovered.

Sorry for the long message but hope it was of some help.

Best Regards,

Jesus Cardin

[joybaggins]

Hello,

Has anyone worked with Xtender before? I have been trying to program the autopilot with Xtender. But there seems to be no documentation that I can refer to. I managed to get it to compile but not sure how to run it or how its supposed to be run.

Appreciate if somebody could help me on this..

Examples will be nice.. =)

Thanks.

[keithskye]

Because of experiences and comments from others such as Jescardin, my team has decided to design and build our own proprietary, integrated flight control computer/autopilot system, that can operate and control a large number of servos/channels on a complex, high-speed aircraft. The system will be somewhat scalable, and it's simplest iteration should be able to fit into a larger RC model, though a full function version will need a bit more room for all the sensors and other periphereal equipment necessary for full auto takeoff-and-land operations (operating multi-servo-redundant elevators, ailerons, rudders, flaps, spoilers, leading edge devices and controlling not only landing gear and throttles, but also mixture settings on IC engines or auto-ignition on turbine engines). We also plan to have auto-braking. Many of these functions are part of the "flight control computer" as opposed to the actual autopilot.

[hwl]

Since this thread pops up from time to time I think it would be appropriate for me to address some of these issues identified in these posts.

First, as of this post, MicroPilot has been selling UAV autopilots for over 15 years. In that time our products and our processes have evolved and improved tremendously. We are ISO9001 certified and almost 750 different organizations have used our autopilots in a variety of different airframes. While we cannot say we have flown every type of airframe imaginable we have flown almost every type of airframe. We are continuously improving our quality control processes. We have an automated calibration process that we use to setup, calibrate and test our autopilots. Each autopilot then goes through a hand test of critical values and as a final check we randomly test fly about 15% of our autopilots. So far we have never encounter a failure in one of these test flights. We also have a number of customers who produce UAVs in quantity and it is rare that these customers encounter a problem with our products.

The company that Jescardin used to work for integrated our MP2000 autopilot into four different airframes – all of which are complex. We haven’t sold the MP2000 for over 5 years and the software that Jescardin was using would have dated from about 7 years ago. Our software has advanced tremendously since that time in terms of its capabilities and its ease of use. The issue that Jescardin described was traced to an installation error that resulted in a blocked static source on the airspeed sensor. As the UAV climbed the pressure on the static side of the airspeed sensor increased which caused the airspeed to read low. Once the customer reported the problem to our technical support it took about a week to resolve. This customer still buys plenty of autopilots from us and it is extremely rare that they experience a failure. Jescadin is not correct in his assertion that our autopilots will not automatically trim a UAV. The I term on a PID feedback loop performs a trimming function. So long as your autopilot software turns off your I terms when you are on the ground, automatically adjusting the trim is quite easy.

Certainly an enclosed autopilot is easier to work with when you are producing a small number of UAVs and in 2006 we didn’t have an enclosed autopilot. However, if you are producing UAVs in quantities it quickly becomes more effective to design a custom circuit board that reflects your particular UAVs requirements and then plug a "board" autopilot into this custom circuit board. This will simplify your wiring harness significantly, reduce your assembly time and cost and increase your reliability. If your autopilot vendor does not sell "board" autopilots this option is not open to you. In addition, you cannot use an integrated autopilot in very small UAVs such as IAI’s mosquito (which uses our MP2128).

It is also true that, in 2006, I/O was difficult to manage with the MP2028/MP2128. At that time our development efforts were focused on the heli version of our autopilot. We have since improved the I/O capabilities of all our autopilots so that you can configure I/O pins for input/output/PWM in/PWM out/serial in/serial out. We also offer remote serial ports on our autopilots. We released our heli autopilot almost two years ago and have shipped a couple hundred since that time. Neither of our North American competitors has released a heli version of their autopilot (although one sells their autopilot hardware with someone else’s software but the price is quite high).

Finally, with regards to keithskye’s announcement, welcome to the industry. Selling UAV autopilots is a great business but it is not an simple business. There are many systems that you need to produce reliable autopilots. It’s useful to be ISO9001 certified to make sure that your processes are properly documented and your documents are properly managed. The experience of dealing with the needs of nearly 750 clients is very helpful. You need good version control and issue tracking. A nightly build and automated regression test really helps catch errors (but it’s a lot of work to implement and maintain). Static code analysis tools are very useful but expensive. An automated calibration system is also important – its OK to have an engineer hand calibrate your autopilots when you sell a couple here and there but you can’t do that when you get a quantity order.

Best Regards,
Howard