This may have been discussed or surely thought of before but Ive not come across it on a thread as yet so Im going to put it out here.. im lucky enough to work with most of the latest technology available to us in this multirotor industry. What keeps coming to my mind is flight controllers and stabiliser gain settings and the amount of time it can take to set them up right...why cant these controllers do a their own flight gains settings, why cant they simply be put on the copter, we input the basic settings ie rotor config, failsafes. But for flight gains... simply take the machine out for a flight in which the controller does its own pre determined in flight checks/flight, using it sensors and gyros surely it must be able to workout what the copter is doing and so set the best gain possible for stable flight and even make allowance for different weights been carried. Does this sound crazy or is this maybe something of the future?
Just a thought on the future of Flight Controllers??
- Thread starter eyeball
- Start date
I would not think that idea is crazy at all. However, these are very complicated pieces of engineering. The hardware portion, accelerometers etc. are the simple part. The software is where the devil in the details is. If the hardware cannot provide the software precise and valuable information, the software cannot make the necessary calculations.
My thought is something like the fly by wire systems found in most modern jets. Usually a trio of flight computers work together to evaluate the pilot's input and take the necessary readings from the sensory input. In a normal circumstance, all 3 would provide the same information based on the criteria. However, if there is some deviation, 2 systems would provide information that does not match and then the third would provide that input for necessary flight control. I know the US's F-117 was one of the first airfames to have that type of setup. Although the expense is probably prohibitive for us it could be the future of our environment.
That or someone just builds a bulletproof system, which I fully support. Haha.
My thought is something like the fly by wire systems found in most modern jets. Usually a trio of flight computers work together to evaluate the pilot's input and take the necessary readings from the sensory input. In a normal circumstance, all 3 would provide the same information based on the criteria. However, if there is some deviation, 2 systems would provide information that does not match and then the third would provide that input for necessary flight control. I know the US's F-117 was one of the first airfames to have that type of setup. Although the expense is probably prohibitive for us it could be the future of our environment.
That or someone just builds a bulletproof system, which I fully support. Haha.
Open Pilot had something close when they had auto tune in a previous version of the firmware. They no longer have that option but Tau Labs do with their firmware for the CC3D boads.
Hoverfly have one gain setting to adjust and no need to mess around with PID tuning. It is possible to set this up on your radio so that you can adjust the gain via the radio while in flight to adjust for different wind conditions etc.
User Brad Quick on the "other" forum has done this in his BradWii open source code offering. It runs on MultiWii boards, has a following, and apparently the auto-tune works quite well. Personally, I don't want the EFC changing things behind my back, *I'll* make the choice as to what constitues optimum flight and adjust the parameters myself. The coders idea of what is optimal and mine may be quite diffferent.
Bartman
Welcome to MultiRotorForums.com!!
like stacky said, hoverfly has one "gain" to tune but with Mikrokopter I've never had to tune anything, the stock settings have worked great across a wide variety of frames/weights and have handled winds gusting to 35 mph with ease. i'm actually working on a new XY8 with 6S power, standard ESC's, Mikrokopter control, and a Hoverfly GIMBAL for the camera control. what's old is new again. as i said in another thread, MK, IMHO, is the current most stable platform. I've lost track of where Hoverfly is at the moment and the chinese products won't appeal to me until they can do a better job of proofing their firmware updates before release. Hopefully Xaircraft can do better than DJI has been doing.
Last edited by a moderator:
gtranquilla
RadioActive
PID loop control has been common to industry for many years. Foxboro developed a self-tuning PID controller many years ago. It was advertised to have AI - artificial intelligence not long after the term first began to appear. The fundamental principle is that a PID loop controller is optimized when it achieves 1/4 wave damping. Figure 1 in the attached hyperlink http://h240.marcks.cc/downloads/10_tuning.pdf document shows the optimum PID tuning that is possible, i.e., 1/4 wave damping. However this is all based on a predictable process such as flow through a pipe where the rate of change of flow changes slowly over time. Not sure if it will work here........ However there are more sophisticated and faster control systems used by the armed forces, e.g., rockets that intercept incoming missiles...... where the need for fastest possible response time is required..... PID loop control is just too slow!
Bart, I came close to heading down the MK route when I first got started but got put off by what I perceived as a technically tricky learning curve. I have been lucky enough to get to fly a MK multi once and it was really nice and I have always had the impression its the gold standard. I havent seen much of late but their FC must be 4-5 years old by now and Im wondering if they have new hardware on the horizon?. However if the processing power of the current FC is good enough it just comes down to firmware really. Although when DJI came out with FC's enclosed in protective casings they did show what MK and HF need to think about with whatever their next gen FC boards need.
DucktileMedia
Drone Enthusiast
Is Hoverfly still in business? I havent heard a single thing from them in ages. Hell, is MK in business??? Maybe they are the one hit wonders of the industry. I'm still sitting back and waiting until something keeps all these numbskulls from crashing their phantoms into people's faces!
R_Lefebvre
Arducopter Developer
Short answer to the OP.
Auto PID tuning is VERY difficult to do. Actually, it's easy to do badly, but it's hard well. Most auto-pid tuning algorithms get things horribly wrong pretty often. It doesn't help that it's a very complicated system we are controlling, and it's not fully modelled.
Also, classic PID tuning doesn't really apply for most our uses. Typically, you would strive for maximum response, with a little bit of overshoot being acceptable. We do not want any overshoot for AP. You want to snug-up to the target, rather than reach it quick, overshoot, and snap back.
There's so much requirement for "feel" that it makes it that much worse.
Auto PID tuning is VERY difficult to do. Actually, it's easy to do badly, but it's hard well. Most auto-pid tuning algorithms get things horribly wrong pretty often. It doesn't help that it's a very complicated system we are controlling, and it's not fully modelled.
Also, classic PID tuning doesn't really apply for most our uses. Typically, you would strive for maximum response, with a little bit of overshoot being acceptable. We do not want any overshoot for AP. You want to snug-up to the target, rather than reach it quick, overshoot, and snap back.
There's so much requirement for "feel" that it makes it that much worse.
Auto PID tuning isn't that hard if you really understand how PID works and mimic what a good human tuner would do manually. The key concept is that you need to be able to separate the effects of the various components (P, I , and D). I posted the theory behind my auto tuning algorithm here:
http://www.rcgroups.com/forums/showthread.php?t=1922423
http://www.rcgroups.com/forums/showthread.php?t=1922423