|
[DPRG] Off Roadbots
Subject: [DPRG] Off Roadbots
From: Randy M. Dumse
rmd at newmicros.com
Date: Sun Oct 29 16:45:34 CST 2006
> It's like your hand on the steering wheel. The car looks like
> it is going in a straight line, but you are actually making
> small adjustments all the time, based on a vector from the
> robot's inertial measurement unit.
Interesting. So you are driving straight by trying to hold the
IMU vector constant? You get a current heading from the IMU, and
difference it with the initial one, and then correct a little
left, or a little right based on that error? Is it a
proportional response to the error, or a higher order? PI PD or
PID?
And what about SR04? You didn't have a IMU then. So weren't you
responding to your odometry?
I suggest that as being your approach, because that is the way I
steer for a point. At first I tried to apply PID to both wheels,
but even tightly controlled, it is not enough. To my surprise,
what the PID would not do (holding the two motors so they go
straight), the odometry was about to observe. While the PID
would detect one wheel falling behind, the correction (I-term)
would again bring to parallel, however if the other side never
lagged, the result was a drifting off zigzag toward the more
burdened side.
I added a proportional response to how the odometry read my
progress. When I added a Ksteer term, and slightly slowed the
inside wheel, that's when driving to a point came together for
me. That's what I meant it wasn't enough to try to go straight,
but one had to steer to go straight.
Sounds like maybe there are other approaches to this, if I
understand your approach correctly.
> Just a test of all the basic robot systems. The more subtle
> navigation improvements come later with exercise #2.
Of course many of these problems will show them self in a "quick
trip" sort of course, and particularly in a long, outdoors,
"quick trip", there are still builders who think locking the two
wheels together is the way to get the robot to go straight.
Randy
More information about the DPRG mailing list
|
