Subject:
|
Re: Detecting tilt with an Accelerometer.
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Mon, 21 Oct 2002 20:15:26 GMT
|
Original-From:
|
T. Alexander Popiel <popiel@wolfskeep.com/avoidspam/>
|
Viewed:
|
1039 times
|
| |
|
|
In message: <H4CIH0.Mz3@lugnet.com>
"Steve Hassenplug" <hassenplug@mail.com> writes:
> In lugnet.robotics, Ralph Hempel writes:
> > > A typical micromachined accelerometer WILL measure tilt of a platform that
> > > is not otherwise accelerating. The value the sensor returns will be
> > > g*sin(tilt angle) if the sensor is placed level on/in the platform.
> >
> > <sound of Ralph eating his poorly thought out words>
> >
> > John,
> >
> > You are, of course, correct. The beam is always under 1 G if it is parallel
> > with the Earth's surface. In free-fall, the beam becomes "weightless" and
> > the accelerometer returns 0.
> >
> > As you tilt the sensor, the beam deflects less and less and eventually
> > there is no strain and it returns 0 as well.
>
> Stop eating for just a sec, and help me understand this.
>
> If a sensor (accelerometer) is parallel to the earth's surface, the reading
> should be zero, right? (no acceleration due to gravity)
No acceleration along the sensed axis, yes.
> But, if the object holding the sensor begins to fall (ignoring the push to
> start the fall), then an accelerometer placed at the center of the object's
> mass will begin to rotate, and accelerate until the rotation is 90 degrees,
> and the acceleration = 1g. At this point the sensor is in free fall, and
> the reading would (still/again?) be zero.
At this point, we need to know the type of sensor you're using
(is it a pendulum type sensor, or a linear compression sensor,
or gyroscopic precession sensor, or something even stranger?),
and the geometry of the fall...
I'm going to assume for the sake of simplicity that you're using
a linear compression sensor. No real reason other than I can
visualize the mechanics of those easiest.
Now for the geometry of the fall. I'll assume that you've got
a two-wheeled inverted pendulum like the LegWay for the robot.
I'll further assume that the sensor is attached to the robot in
the middle (as close to center of gravity as possible), in a
fixed orientation vs. the RCX (the axis of the sensor in the
plane of the wheels, and parallel to the ground when the robot
is upright). Finally, I'll assume that the fall you're talking
about is the robot toppling over, instead of rolling off the edge
of a table and falling down without the body rotating.
So, the fall takes the sensor from a horizontal position to
a vertical position.
Looking at just the gravitational component of the acceleration,
at the beginning of the topple, the sensor is reading zero (all
of gravity is perpendicular to the sensor's axis). At the end
of the topple, the sensor reading is 1g (all of gravity is along
the sensor's axis). The sensor does not reset to zero, it stays
at 1g.
During the topple, the sensor value increases proportionally to
sin(theta), where theta is the angle that the robot body (and
thus the sensor) is tilted away from horizontal.
If, on the other hand, the robot just drove off the table while
remaining upright, then the sensor remains at zero throughout.
Thirdly, if the robot drove off the table and tumbled, then during
the fall a sensor at the center of mass would read zero (regardless
of angle, since the bot is in free-fall), but sensors elsewhere in
the body could detect the centripetal accelerations of the tumble
if oriented radially (and if they're sensitive enough, given that
you're probably talking milligravities at most). Once the robot
landed on the floor, you'd be able to detect angle vs. gravity on
all sensors, again.
- Alex
|
|
Message is in Reply To:
 | | Re: Detecting tilt with an Accelerometer.
|
| (...) Stop eating for just a sec, and help me understand this. If a sensor (accelerometer) is parallel to the earth's surface, the reading should be zero, right? (no acceleration due to gravity) But, if the object holding the sensor begins to fall (...) (22 years ago, 21-Oct-02, to lugnet.robotics)
|
14 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
