Hi Dave,

On Tue, Mar 4, 2008 at 12:05 PM, Dave Tweed <pic@dtweed.com> wrote:
> Sean Breheny <shb7@cornell.edu> wrote:
>  > Thanks! I do wish they had discussed more of the physics, though.
>  > There are a lot of misconceptions out there about what accelerometers
>  > can and cannot do (for example, as you said, you need to know
>  > something about the way it is moving or not moving in order to extract
>  > tilt information from the measured accelerations. Actually, what I'd
>  > like to see is a general app note on inertial navigation, covering,
>  > for example:
>
>  That's not an app note, that's a whole shelf full of books!
>

I'd contend that a basic, practical-level intro to it would not be.
You're not going to be good enough to build ICBMs with it, but you can
start playing with INS for robotics-type applications with this level
of knowledge.

>
>  > 1) That accelerometers do NOT measure gravity, but rather all the
>  >    forces other than gravity which are applied to an object.
>
>  Huh? The accelerometers I have here measure gravity just fine!
>
>  Perhaps you meant to say that they measure the net sum all of the
>  accelerations of an object, including gravity.
>
>

No. If they measured the net sum of all the accelerations, then they
would measure zero for a stationary object.

If you have an accelerometer in free-fall in a vacuum, accelerating
under the influence of gravity, it measures zero. If you then set it
on a table so that the sum of the forces on it is zero, but there is 1
newton of gravity and 1 newton worth of normal force acting back from
the table, it will measure ONLY the force from the table (1 newton).
This EQUALS gravity because you know that it is sitting on a table and
not being allowed to accelerate. It is only under this assumption
(that it is not accelerating) that you can relate its measurement to
gravity.

>  > 2) Models for the typical noise and offset errors of accelerometers
>  >    and rate gyros.
>
>  Very dependent on the implementation technology.
>

Yes, but most of the technologies available to the hobbyist are
similar (almost all MEMS with similar specs).

>
>  > 3) Euler angles vs. rotation matrices vs. quaternions
>
>  Mainly a question of which works better for the math you need to do in
>  your overall applciation.
>

Yes, I didn't mean that it would tell you which to pick, only give a
quick overview of how to use each and their advantages and
disadvantages.
>
>  > 4) Schuler tuning
>  > 5) Kalman filters
>  > 6) Other types of filters (like sigma point filters)
>  > 7) GPS-INS integration, GPS error models
>
>  Like I said, each one a topic for an entire book, and the choices made
>  depend very strongly on the overall application.
>

I know of a very good practical reference for Kalman filters which is
only about 10 pages long. Sigma point is even easier to implement.

Schuler tuning can be explained (again, at a beginner level) in a page.

GPS sources of error can be modeled easily at a basic level.

Sean


>  For example, I'm working with a client who uses GPS interferometry to
>  measure vehicle absolute attitude (roll, pitch and yaw) at low sample
>  rate, then combines this with relative INS measurements in a Kalman
>  filter to produce an interpolated solution at high sample rate.
>  Fascinating stuff, but he handles all of the theoretical stuff and the
>  application software. I just build the underlying hardware that it runs
>  on.
>
>  -- Dave Tweed
>  --
>
>
> http://www.piclist.com PIC/SX FAQ & list archive
>  View/change your membership options at
>  http://mailman.mit.edu/mailman/listinfo/piclist
>
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist