Comments on: Quadcopter Prototype using Acc_Gyro and a PIC

By: DroneGeek

DroneGeek — Sat, 17 Sep 2011 21:14:12 +0000

Awesome quadcopter!!, but as anyone have a look at or tried the Arducopter quad which uses the ardupilot mega board? I am interested in getting one but not sure how well it performs?

http://unmannedtechshop.co.uk/Ardupilot-Mega-2560-kit

By: Rafiq

Rafiq — Tue, 09 Aug 2011 13:06:26 +0000

can you give me the source code for PIC programs. I really need it, I decided to make quadcopter for my final project. I need it as soon as possible

By: phoenix

phoenix — Thu, 05 May 2011 07:05:45 +0000

I have the same quadrotor structure and I use pic18f6520 and a 6 dof IMU but it seems impossible to fly. you can see my previous video here:
http://www.youtube.com/watch?v=DIP34T5SFtw&feature=youtube_gdata_player

Although I have managed to get rid of the vibrations I still cannot make it fly. My guess is to change motors to brushless DC and use ESC for speed control. Also I think we should change the blades too.

By: fahdovski fa

fahdovski fa — Tue, 08 Feb 2011 21:19:16 +0000

Thank you Starlino for your article
Very Helpful !!!
Im Building my first quad with
Usb Hid Joystick -> C# Xbee Pic182550
Imu = AdxL330 +Lpr450AL
All card home made

Thanks Starlino

By: starlino

starlino — Thu, 27 Jan 2011 19:02:30 +0000

I am using Rhino 1350mAh 3S 11.1v 25C Lipoly Pack (or something close and similar). I got them from http://hobbyking.com which seems as a good site for all RC parts. If you’re there grab a small camera for your first flight :) . What type of ESC do you need brushless or brushed ? In any case you can ask the veterans on http://www.rcgroups.com I am sure they have plenty to advise. As far as my project I settled on the onboard MOSFET to save costs. Brushless motors will last longer but will not necessarily give you more precise control since you can only update them every 10-15 ms. There are newer I2C brushless ESC though that can be updated more often but they are expensive. Any approach you take will have it’s challenges. I think PIC is great since you get 80Mhz processing and many features that Arduino does not have (yet). I have to agree C18/C30 is harder to program than Arduino but once you get it you will love the extra features and control you get.

By: Brian

Brian — Thu, 27 Jan 2011 18:45:41 +0000

I am following and using this project as a beginning to our own UAV quadcopter project. I am going to use ESCs instead of the FETs and SchKotty diodes. We are using a similar PIC. Most other sites are using the Arduino boards and we are designing our own.

2 Quesitons: What is the best battery to use and what would be the recommended ESC to use as well?

By: Chesley88

Chesley88 — Sat, 11 Dec 2010 08:47:22 +0000

Hi~ guys, currently i headache with programming the pitch and roll axis of quadcopter.

Right now , i fixed the quadcopter on the wood structure for control the front motor and back motor speed , in order to achieve balancing flight.

I’m using Arduino as microcontroller, and 5 DOF IMU (gyro and accelerometer) to determine the estimated angle.

In my programming,i stated that ,if overall buttons are at the low condition, which means all buttons are switch off. Then i using P term for balance the QuadCopter, the programme as below:-

if(digitalRead(UP_button)==HIGH && ButtonRelease1 == true)
{

x=constrain (x+20,1000,1800);// front motor

z=constrain (z+20,1000,1800);// back motor

ButtonRelease1 = false;

}
if(digitalRead(UP_button)==LOW ) ButtonRelease1 = true;

if((digitalRead(UP)==LOW) &&(digitalRead(DOWN)==LOW) &&(digitalRead(FORWARD)==LOW) &&(digitalRead(BACKWARD)==LOW) &&(digitalRead(LEFT)==LOW) &&(digitalRead(RIGHT)==LOW))
{
int Kp; //tuning parameter,Kp
int Pout; // Proportional output,which is P term.
int Sp=0; // Setpoint
int PV ; // process value, angle estimated

Kp=1;
PV = ( Rate_of_Estimated_Angle * 90 );
Pout = ( Sp – PV ) * Kp ;
X=x-Pout; //x is previous front motor speed
//X is the new motor speed
Z=z+Pout; //z is previous back motor speed
//Z is the new motor speed
}
PWM2.writeMicroseconds(X);
PWM4.writeMicroseconds(Z);

But the problem is —– The QuadCopter always fluctuating !!!! Any mistake i did wrong with the code as above ? Thanks for all ur information~

By: Jeroen

Jeroen — Fri, 29 Oct 2010 21:30:40 +0000

How big is your angle noise when quad is still and motor turn. I see about few degrees noise. This discreases when I ingrease wGyro. Also want to try put some white noise on the motors and see what the angle does to determine a model in matlab.

Also bolted the electronics on rubber, but don’t see a big difference in noise.

By: Jeroen

Jeroen — Mon, 25 Oct 2010 10:30:53 +0000

I shall look on the new software, waiting on one new motor a bearing broke due a crash.

Updated my BLDC controller to run on 2kHz controlloop. The acrobat mode runs now on 1kHz not yet tested due the motor. Also orderded some rubber dampers to reduce the vibarations. hopefully I’m have some time to test more this week.
The adaptive wGyro could be a good thing. I will keep you updated

By: starlino

starlino — Sun, 24 Oct 2010 17:07:42 +0000

Jeroen , you should not disregard the fact that you can improve mechanical factor (make sure weight is balanced, introduce amortization under sensors foam/tape). For example my quad was drifting yesterday at one side I discovered it is not balanced I repositioned some components to make it more balanced and got better results. I updated the code :
http://code.google.com/p/picquadcontroller/updates/list
There is still a lot of tuning to do, flight is far from stable, but much better, and I am able to “trim” the roll/pitch/yaw command values, the params I use with best results are Kp(Roll&Pitch) = 25 , Kd(Roll&Pitch) = 200 , Kd(Yaw) = 80
I can see it starts oscillating if I increase Kp(Roll&Pitch) since this signal is delayed and not very reliable when quad is subject to external accelerations, I modified the imu.h code to observe if modulus of acceleration is far from 1g to give more weight to gyro and less weight to accerlerometer. I actually replaced wGyro with accWeight as you’ll see in the code but that’s equivalent.

By: Jeroen

Jeroen — Fri, 22 Oct 2010 07:11:49 +0000

that is what I thought, I tried several values ranging from 5 to 20. wGyro=20 gives a nice smooth signal but the quad starts osscillating at a lower gain. I think this caused by the phase delay of the filter. Also played with the internal filters in my gyro and acc. Set my gyro to 188Hz low pass filter which works fine in angle velocity control mode. So let the gyro the same. The Acc has also an internal filter ranging from 1200Hz to 10Hz cutoff freq setpoints. If I also lower this value to 40Hz it also starts oscillation faster.

I do see alot of noise on the angles when the motor is on. This can be reduced with the internal filters or wGyro but the quad will start oscillating. All ready posted on RCgroups but don’t get much response.

Tonight will also try to get the tranfer model of my motors so I can tune them better. Matlab is a great tool for tuning.

By: starlino

starlino — Fri, 22 Oct 2010 00:44:34 +0000

Jeroen, it is so much better now , I didn’t do an in depth analysis of the frequency response , all I know it is an IIR (Infinite impulse response) low pass filter, the cutoff frequency is determined by wGyro , which I determined experimentally by collecting data with motors on. Please note there’s yet another filter in my code it’s a simple averaging filter in the adcutil.h module , it collects 16 samples to the DMA buffer and then averages them for the loop interval mine is around 6 ms since I use direct PWM to motors not servo controllers which requires at leas 10-15 ms to work properly.

By: Jeroen

Jeroen — Thu, 21 Oct 2010 18:06:40 +0000

I now got the angle velocity control pretty stable
http://www.youtube.com/watch?v=h9_zp5WW4jo
This is ony 200Hz PID controlled loop (now able to run on 1kHz).

I can’t get the angle control stable, deleted my angle filter and implanted your seems to work good. But cant get the control loop stable. I logged some data form the motor control value and the angle when is osscillation and used matlab to determine a model but this fits reality. Did this also with angle velocity control loop that works perfect.

Do you no how what the frequency response of your filter looks like ?

By: Jeroen

Jeroen — Wed, 13 Oct 2010 17:58:04 +0000

I tried holding the sensor with my finger (still in pinheader socket) and see no difference.

Here is a video of the output of the gyro white line deg/sec:
http://www.youtube.com/watch?v=dzetSxw3S-Q

I just found out setting the bandwith in the gyro to a high level gives less drift 1kHz sampling 100Hz bw is worse than 8kHz 256Hz bw with much less phase delay. This is because of the higher bandwith I could set P-gain to 22 instead of 15 which seems to help. Still a bit jittery but this could go away with tunig.

To bad my battery is empty and only have one.

By: starlino

starlino — Wed, 13 Oct 2010 13:44:38 +0000

Jeroen , how about a mechanical dampening of the sensors, after I mounted my sensors using foam tape instead of inserted in socket, I got much less noise.

By: Jeroen

Jeroen — Wed, 13 Oct 2010 08:18:51 +0000

I have tried I term but it doens’t seem to help much (just a little bit). I made a better test stand so the quad can rotate on one axis at full speed (2000deg/sec) with this I can determine the step response and make a model for matlab. Which makes is easier to tune. I have done this yesterday but I only could measure 1sec which it not enough.

From the results of yesterday I have seen that there are high frequence noise when the quad is rotating. So I have to look to the optimal setting for the internal filters without to much phase delay.

By: starlino

starlino — Tue, 12 Oct 2010 23:26:31 +0000

Ohh I think you should also consider introducing the I (integral of errors) , this looks like a prolonged drift that is exactly what I factor fixes. Find a good resource of PID tuning and try all Kp, Kd, Ki factors I think you’re close. With integral you have to make sure it doesn’t go crazy (so you have to limit how much it can grow, reset or trim it from time to time).

By: Jeroen

Jeroen — Tue, 12 Oct 2010 18:49:06 +0000

http://www.youtube.com/watch?v=KZQXdGFWGXE

In this movie you can see the drift

here is an excel sheet with step reponses of one axis with a sample time of 1ms.
http://users.telenet.be/CyrusTheVirus/response.xls

By: starlino

starlino — Tue, 12 Oct 2010 17:34:07 +0000

looks really good, I am now behind you on this project :) have you thought of recording the readings from sensors/control wirelessly and then mapping it on the video (you can use led to synchronize start of recording), that way it would be easier to debug those drifts. it’s hard to tell at this point what is wrong , can be more than one thing.

By: Jeroen

Jeroen — Tue, 12 Oct 2010 11:50:08 +0000

I now made a PD control on the angluar velocity which seem te work with little drift:

http://www.youtube.com/watch?v=lfPuM1grSLQ

But the absolute angle control won’t be stable with PID controller. What could it be the angle calculation or something else (2 posts above). With a P-gain of 10 is starts oscillation and it isn’t still at all.