DPRG List  

[DPRG] April 2009 Meeting Minutes & PID and Beyond Presentation Summary

Subject: [DPRG] April 2009 Meeting Minutes & PID and Beyond Presentation Summary
From: DeltaGraph at aol.com DeltaGraph at aol.com
Date: Mon Apr 13 04:30:28 CDT 2009

DPRG Monthly Meeting Minutes 4/11/09
Present: Ron Grant, Paul Bouchier, John Swindle, Will Kuhnle, Brian Pike,  
Drew Dolan, Ed Paradis, Glenn Pipe
There was no March meeting - the club attended All-Con instead.
- Ron showed his audio-output from motor ramp generator, and his  robot 
which lets you play with PID parameters
- Ron described how Doug  brought the club $1500 (from Verizon donations), 
and he asked people to think  about whether they would support the idea of 
approving Doug's desire to guide  it's application. Doug is excited about 
YARB, and would like the club to buy  one.
- Ron described YARB (Yet Another Robotic Blimp) - $350 for blimp OR  $700 
for blimp & blackfin robot + camera (discount from $900)
- There's a  Blimpduino - around $100, but doesn't carry much payload
- Paul suggested  maybe the club should perhaps auction off the robomower 
to offset blimp cost.  Discussion of need to find out whether the donor is 
amenable to that kind of use  of the donation. We need to get in touch with 
the donor. Question: what laws or  rules apply to who can participate in such 
an auction. Paul got action to find  out from Steve Rainwater who donated 
it, and what rules apply to any such  disposal (e.g. 501c rules etc).
- Ed mentioned Akon - 2 hour  presentation
- Ron said Tanner Electronics wants DPRG to do a robot day in  October or 
November - dual purpose - people  can shop & support the  demo. Tanner has 
offered to let us show DPRG info on the wall. People liked both  of those 
- Ron won't be here next month's meeting, Glenn  (Vice-President) is 
responsible for organizing next month's meeting.
- Will  presented a talk on PID (proportional, integral, derivative)
A presentation on control systems, by  Will Kuhnle
[Note: We Expect to Have Will's Slides Published to DPRG.ORG]
- Will handed out notes for the talk, & pointed out: if K & G are  large, 
the output essentially becomes equal to the output, because KG / KG  
approaches 1
- As an example, the F16 is unstable without feedback  control
- Feedback permits greater tolerance on control components G
- The  dotted line in the speed/torque graph shows that as load increases 
speed  decreases
- Servos may not compensate for noise in the system - e.g. bumps on  the 
- Will showed his propeller-driven pendulum.The axes of the  feedback time 
delay graph are: X: ms, Y: voltage applied to propeller. He found  that 
thrust is almost linear with current, but speed is not.
- The voltage  constant & torque constant are related. The time constant of 
the  motor/propeller is around 50 - 60ms
- Will emphasized the importance of  thinking about what you're trying to 
actually control, when designing a control  system. Also, you can't control 
what you can't measure - be sure you can measure  what you're trying to 
control, and ask is it the right variable to control, to  meet your goals?
- You may want to switch what you're controlling over time.  E.g. when 
sending a robot to a target, you may want to control speed for much of  the 
travel, so as not to spin the wheels due to saturated control system, but  when 
the distance becomes less than some value, you may want to switch to  
controlling distance to the target. Will's list of variables that can be  
controlled on a 2-wheel driven robot shows the wide range of variables that can  be 
controlled, perhaps at different points in time. It really is important to  
think about what you want to control.
- Philosophical question from Ed: is  there some physical reality to the 
integral of distance with respect to time? -  Unanswered.
- Will showed that a motor that's part of a position servo can be  modeled 
as a mass on a spring, which will oscillate in the absence of friction.  The 
derivative term acts like friction on the mass + spring assembly.
- Will  explained why you need the integral term, by using an analogy with 
2 tanks of  water connected by a pump that pumps water from tank 1 to tank 
2. If you're  trying to control the level of tank 2, the pump could simply 
run until the water  is at the right level. But now if you start draining 
water out of tank 2 at a  steady rate, there will be an error in the desired 
level, because the pump won't  run until there is an error, and the error will 
remain, as long as you're  draining water out of tank 2. Anytime you have to 
supply power to hold the  desired value, you need integral, otherwise 
you'll have a constant error.
-  On a position servo, you need derivative control, to stop it from 
oscillating.  But on a speed controlled servo, you don't any derivative, because 
when you get  to the desired speed, you can just stop applying power. This is 
because there's  only a single integral in the controlled system - it 
doesn't have "momentum"  that will cause it to tend to overshoot in speed. So 
it's really important in  robotics to know what you're trying to control - if 
you're controlling speed,  you don't need any derivative, but if you're 
trying to control position, you  do.
- Discussion of a bode plot - it's the output of the process (e.g.  
position of a motor) with the control-loop feedback removed, when the input is  
driven by a sinusoid. The point where the motor starts to fall behind the input 
 sinusoid is the time constant of the motor - typically ~30 - 60ms for 
small  motors. You need to have a gain of less than 1 at the point where the 
frequency  response on the bode plot drops from flat to the 20dB/octave falloff.
- Servo  motors are designed to have a high torque to inertia ration - this 
is why they  tend to be long skinny motors to keep inertia low (because 
inertia varies as the  square of diameter of the rotor).
- When you're changing the amount of  integral or derivative you're 
changing the frequency on the Bode plot at which  the response changes.
- New techniques: There's an alternate approach to  control systems from 
PID - the State Space approach. PID is a technician's  approach - tune it 
until it works. State space is a different approach to how  you design a servo 
system.  You feed in position, velocity, acceleration,  jerk in different 
amounts. PID is a mere combination of the derivatives. In  state space, you 
have a matrix of variables, but we didn't get a real good  explanation of it.
- There's a difference between state space & Kalman  filter - in a kalman 
filter you can't measure all the states, so you come up  with a math model of 
the system. If you have a perfect model, you can get your  states out of 
the model. But your model isn't going to be perfect. So you  measure system 
output & feed it back into the math model as a correction. A  Kalman filter is 
a state space estimator.
- Discussion of types of system.  System type is the number of pure 
integrators in the system: a type 1 system has  1 integrator.

**************The Average US Credit Score is 692. See Yours in Just 2 Easy 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.dprg.org/pipermail/dprglist/attachments/20090413/6326f979/attachment.html

More information about the DPRG mailing list