Searching \ for '[PIC]: speed from acceleration' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/microchip/devices.htm?key=pic
Search entire site for: 'speed from acceleration'.

Exact match. Not showing close matches.
PICList Thread
'[PIC]: speed from acceleration'
2003\06\18@184946 by Tony Nixon

flavicon
picon face
Hi all,

I would like to make a speedometer for a vehicle but would like to make
one that needs no calibration (per say).

As the vehicle will not be travelling for long times, I am thinking of
using one of the ADXL digital accellerometers and try to calculate the
speed from the acceleration obtained.

Is this achievable and if so, does anyone have an idea or links to the
math involved.

best regards

Tony

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@190346 by Ted Larson

flavicon
face
Tony,

I have been fooling with the new Memsic accelerometer.  It has digital
output...so you don't have to do a bunch of A/D in order to read it.
And...its cheaper than the ADXL, and you can buy a version that has
auto-thermal calibration, so you don't have to calibrate it yourself. They
have an app note on their website that describes the exact application you
are trying to do.

http://www.memsic.com

Hope this helps,

- Ted



{Original Message removed}

2003\06\18@192034 by Bob Axtell

face picon face
We designed a protoctype one for NASCAR use; it worked by counting pulses
from a magnet mounted on a rear axle, and was calibrated by the driving the
car exactly one mile manually, and the uP handled all the math and display.

But the problem turned out to be errors caused by tire wear; the same
number of turns no longer meant the same distance. Worse, it happened in
the middle of the race! So it was not adopted, although it could be
recalibrated by an IR burst every km or so, at little expense.

The fun part of that job was meeting the folks in Pit Row.

I think it would be tough to do it by accelleration alone; they do that in
space because there is no wind resistance or other factors, such as moving
up and down.  They then recalibrate frequently with star sightings.

--Bob

At 08:48 AM 6/19/2003 +1000, you wrote:
{Quote hidden}

---------------
NOTICE

1. This account can accept email & attachments up to 10M in size.
2. Federal Monitors: At request of client, some attachments are encrypted.
Please DO NOT delay traffic; please reply with credentials for password.
--------------

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@193931 by Adam Nelson

flavicon
face
Ted:
As it happens, I am also interested in approximating speed from
acceleration, over short distances.

I am probably missing something profoundly obvious, but I could not find
an app note on MEMSIC's site that addressed this issue.  Is it
non-obvious, or am I an idiot?

Thanks,
Adam


> {Original Message removed}

2003\06\18@193939 by Spehro Pefhany

picon face
At 08:48 AM 6/19/2003 +1000, you wrote:
>Hi all,
>
>I would like to make a speedometer for a vehicle but would like to make
>one that needs no calibration (per say).
>
>As the vehicle will not be travelling for long times, I am thinking of
>using one of the ADXL digital accellerometers and try to calculate the
>speed from the acceleration obtained.
>
>Is this achievable and if so, does anyone have an idea or links to the
>math involved.

It's just integration, no big deal mathematically. The problem is that
the DC error in the accelerometer will give extreme levels of
inaccuracy and drift unless:

1)      The zero speed condition can be detected reliably

2)      The acceleration is always robust compared to the
        full scale range of the accelerometer

3)      It doesn't sit for very long at speed


Depending on what you are doing , you might want to look at the new
MEMS dual axis gyro chips. Might be nice (if that's the right word)
for things like guided artillery shells.

Best regards,

Spehro Pefhany --"it's the network..."            "The Journey is the reward"
spam_OUTspeffTakeThisOuTspaminterlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@195445 by Tony Nixon

flavicon
picon face
Bob Axtell wrote:
>
> We designed a protoctype one for NASCAR use; it worked by counting pulses
> from a magnet mounted on a rear axle, and was calibrated by the driving the
> car exactly one mile manually, and the uP handled all the math and display.
>
> But the problem turned out to be errors caused by tire wear; the same
> number of turns no longer meant the same distance. Worse, it happened in
> the middle of the race! So it was not adopted, although it could be
> recalibrated by an IR burst every km or so, at little expense.
>
> The fun part of that job was meeting the folks in Pit Row.
>
> I think it would be tough to do it by accelleration alone; they do that in
> space because there is no wind resistance or other factors, such as moving
> up and down.  They then recalibrate frequently with star sightings.
>
> --Bob

We do similar to that now, by using a gear tooth sensor mounted close to
the punched slots in the car wheels. Just drive it steady at speed (X)
and the PIC can calibrate itself. But as we never know what type of car
we will use, it's a pain and very time consuming to set up and calibrate
each time.

I just found this device that does the job, so it seems it can be
accomplished, but expensive in this case.

http://www.vericomcomputers.com/

regards

Tony

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@195859 by Picdude

flavicon
face
Tony,

I've looked into this, and although accuracy is not an issue, it will eventually "accumulate" to the point where the vehicle may read a couple miles per hour when parked.  There would still need to be some other signal from the physical movement of the vehicle to reset it when stopped.  I had looked at sensors from Measurement Specialties, but until I find a reasonable way to do this, the idea/project has been put aside.  Any thing I could think of would require some sensor connected to the transmission or axle, and at this point, we might as well just use the regular speed-sensor method.

The G-Tech units do this for 1/4-mile-time calcs and get pretty good accuracy, including speed, but it needs to be "reset" before each run.

Math should be very easy, going back to the basic motion formulae, v=u+at, s=ut+1/2at^2, etc.  [ Man, it's been a while! ]

Cheers,
-Neil.



On Wednesday 18 June 2003 17:48, Tony Nixon scribbled:
{Quote hidden}

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@200710 by Picdude

flavicon
face
On Wednesday 18 June 2003 18:19, Bob Axtell scribbled:
> But the problem turned out to be errors caused by tire wear; the same
> number of turns no longer meant the same distance. Worse, it happened in
> the middle of the race! So it was not adopted, although it could be
> recalibrated by an IR burst every km or so, at little expense.

Estimating and adjusting for tire wear should not be a problem, and I'm sure it can be done quite accurately.  It's the calculations for diameter-increase of tires due to centrifugal force that really add to the fun.  Again, it can be calculated for, *if* this parameter is known/measured.  Ugh!

An accelerometer does shine in this case, but a speed (or at least moving/standstill) sensor would still be needed to keep it in check, probably during standstill.  But at this point, I'd think GPS starts looking like a more attractive solution.

Cheers,
-Neil.

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@201542 by Picdude

flavicon
face
On Wednesday 18 June 2003 18:39, Tony Nixon scribbled:
> I just found this device that does the job, so it seems it can be
> accomplished, but expensive in this case.
>
> http://www.vericomcomputers.com/

The Vericoms were used by the auto magazines to performance-test vehicles, with the external wheel attached.  The G-tech came out with the accelerometer version, which did the same thing, with excellent accuracy, and was tons cheaper.  Vericom has been playing catch-up, but not price wise yet.  Here's a link to the G-tech device...
  http://shop.store.yahoo.com/gofast/gtecgtcombrp.html

http://www.gtechpro.com is their site, but seems to be down now.

Cheers,
-Neil.

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@223607 by Marc Nicholas

flavicon
face
On 18/6/03 20:07, "Picdude" <.....picdudeKILLspamspam@spam@NARWANI.ORG> wrote:

> An accelerometer does shine in this case, but a speed (or at least
> moving/standstill) sensor would still be needed to keep it in check, probably
> during standstill.  But at this point, I'd think GPS starts looking like a
> more attractive solution.

Bounce IR off the asphalt and then detect the scatter pattern. Vaguely
unified signal == not moving. Combine with an accelerometer and you've got a
good dead-reckoning system.

Don't be looking at GPS!


-marc



--------------------------------------------------
Marc Nicholas Geekythings Inc. C/416.543.4896
UNIX, Database, Security and Networking Consulting

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\18@234320 by Robert Rolf

picon face
Spehro Pefhany wrote:
>
> At 08:48 AM 6/19/2003 +1000, you wrote:
> >Hi all,
> >
> >I would like to make a speedometer for a vehicle but would like to make
> >one that needs no calibration (per say).

Well, a GPS receiver will get you that. And you get direction
and distance as a bonus <G>. GPS chip sets are getting to be really cheap
and since most racing takes place outdoors...


> >As the vehicle will not be travelling for long times, I am thinking of
> >using one of the ADXL digital accellerometers and try to calculate the
> >speed from the acceleration obtained.

And how will you handle the suspension tilt that results from acceleration
and braking? Or bumps?

{Quote hidden}

4) The acceleration is completing in the axis you are measuring.

> Depending on what you are doing , you might want to look at the new
> MEMS dual axis gyro chips. Might be nice (if that's the right word)
> for things like guided artillery shells.

Again, DC drift is an issue.

Robert

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.

2003\06\19@004924 by Robert Ussery

flavicon
face
This is also known as inertial navigation, widely used in airliners and
cruise missiles. With enough money and time, very good accuracy is possible.
By using lots of laser gyros and complex computing, the airliners manage
very good accuracy.

- Robert :O)

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email listservspamKILLspammitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@013324 by Ted Larson

flavicon
face
Adam,

Whoops..I was mistaken...The paper I was referring to is relevant....but,
not EXACTLY the solution you are looking for as I had said previously.  It
is related to sensing inclination angle of a moving vehicle, which could be
used for dead reckoning during GPS signal loss.

http://www.memsic.com/memsic/pdfs/an-00mx-012.pdf

Interesting nontheless.  Perhaps helpful.  You could probably combine the
information with the output of a rate gyro, and use it to solve your
problem.  A small piezo gyro like they use for image stabilization in a
camcorder would work.  Check this out...small..and cheap too.
Look at the ceramic gyro on page 23 of this document:
http://www.nec-tokin.net/now/english/product/pdf_dl/sensors.pdf

Hope this helps,

- Ted


{Original Message removed}

2003\06\19@021555 by Trevor Page

flavicon
face
Hi,

My friend and I are madly into our modified VW's and we have been
talking about making our own little performance monitoring unit for a
long time. We were going to make one that had a graphic LCD so that it
could actually show your 1/4 distance / speed / acceleration graphs on
the screen for you. Then, I noticed that there is a commercial device
that already does this - the G-tech - and I always intended to find out
more about it. It's interesting to hear it mentioned in this thread, and
thanks for the links.

This thread is fascinating for me because I too have been wondering
exactly how accurate it would be if you integrated the data from an
accelerometer in order to get distance / velocity. I know it is utterly
simple maths, but I always wondered how significant the DC offset would
be. From what's been said here, it looks like it gains a reasonably
large relatively quickly.

In order to determine when the vehicle is at rest, how about a vibration
sensor of some kind? The vehicle will always be subjected to vibration
when it is at rest, but surely a lot more so when it is starting to
move. Or maybe the vibrations from the engine itself would 'drown out'
any vibration signals from the actual movement of the car, if the sensor
was mounted in the engine bay. Maybe if you ran the sensor along the car
into the rear somewhere. I know that a sensor on a axle is better, but
what I am trying to think of here is a sensor that is easy to install.

Oh this is my first post on here in a long time by the way. I've just
subscribed back onto the list. Would someone mind telling me what I have
to include in the message for my email address not to be included in the
groups.google etc. archives please?

Cheers

Trevor.

> {Original Message removed}

2003\06\19@022228 by Trevor Page
flavicon
face
And also - sorry if I am being dumb here - but if you had a sensitive
enough accelerometer, surely you could rely upon this giving a stable
output, for more than a few seconds, in order to detect that the vehicle
is at rest, and hence use this signal to reset the speed measurement?
I'm going by the assumption that a moving vehicle will never be at a
constant speed for very long in reality.

Trev

> {Original Message removed}

2003\06\19@023513 by Tal

flavicon
face
Yes, this is achievable. Here is a product that does just that.

http://www.gtechpro.com/gtechprocomp_howitworks.html

I used the older version of that product and was impressed how much they
can do with a simple accelerometer (horse power, skid pad, 0 to 60 time,
etc). Even the data entry used the accelerometer. To enter the car
weight, you tilt the accelerometer up and down and this
increase/decrease the displayed value until you get to the desired
value.

Tal

> {Original Message removed}

2003\06\19@033100 by Alan B. Pearce

face picon face
>Bounce IR off the asphalt and then detect the scatter pattern.
>Vaguely unified signal == not moving. Combine with an accelerometer
>and you've got a good dead-reckoning system.

I believe this is effectively how they measure the speed of steel in a hot
rolling mill, by looking at the pattern of the glow in the steel, and
watching how fast it moves.

However for Tony's purpose I suspect a better bet is to use radar sensing -
think police speed trap. I would start with a microwave person sensor as
used on burglar alarms as something to hack for the RF side. Then you "just"
need a frequency counter to measure the speed. Again could get quite
inaccurate at very low speed, but you cannot have everything. No problem
with top speed or constant speed with this system.

Just what would be the lowest speed that you need to measure?

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email .....listservKILLspamspam.....mitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@033123 by Stephen Webb

flavicon
face
> In order to determine when the vehicle is at rest, how about a vibration

I don't know how these accelerometers work, but I'm thinking that you can
tell when you aren't accelerating when the length of the force vector (3
axis, right?) is qppx equal to 1 g.  (you also get the direction of the
gravity vector that way, which is helpful if you don't mount the thing
perfectly horizontal or whatever).

Or just long periods (1 second, say) where the variance in the
measurements is pretty low. (whatever that means)

-Steve

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email EraseMElistservspam_OUTspamTakeThisOuTmitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@033723 by Hugo Harming

flavicon
The may issue of Circuit Cellar had a remarkably simle G-force meter project.
Simple circuit and simple PICbasic source.
Used the ADXL202 IIRC.

/Hugo




{Original Message removed}

2003\06\19@034421 by Hugo Harming

flavicon
It says the thing picks up engine RPM from the cig. lighter socket! Brilliant!
Anyone done this?
What signal levels to expect?
Is filtering required?


/Hugo (analog stuff scares me!)




{Original Message removed}

2003\06\19@084719 by Dennis J. Murray

flavicon
face
Hey,all!

I'm probably missing something obvious here, so forgive my ignorance.  I've
played some with the Analog Devices ADXL series accelerometers and been
quite impressed with them.  It seems like your idea of deriving speed from
acceleration could be accomplished with a hybrid solution.

Why not use BOTH an accelerometer and a wheel/axle/driveshaft sensor??  Use
the wheel sensor to detect "at rest" condition and very slow acceleration?
You could determine "instantaneous" speed from the ADXL chip and
mathematically back into how many counts that speed represents on the wheel
sensor (i.e. recalibrate the wheel sensor?).  When you have constant speed
OR very slow acceleration, use the recalibrated wheel sensor for that data.
When the acceleration gets high enough to be in the linear (acceptable?)
range of the ADXL, then switch over to that chip for your data.  You
wouldn't have to recalibrate the wheel sensor every reading of the ADXL,
just when the reading dropped below some arbitary acceleration threshold.
None of this is difficult programming, so we're not talking rocket science
here.

No, I have not done this, it's just a suggestion.  Personally, I'd like to
hear how you've solved this problem!

Dennis


{Original Message removed}

2003\06\19@093026 by Mike Hord

picon face
I haven't heard anybody mention the article in Circuit Cellar last month, so
I
will.

Seems a gent made an accelerometer for his new Beetle...Two axes only,
but the process could be expanded.

I don't know if you can do vector math with a PIC, but it seems to me
that the easiest way (if you're hell-bent on doing it G-based) would be to
install the sensor as close to perfectly level as possible, then use vector
sums
to calculate the acceleration.  Bumps and such would filter themselves out
because what you gain on the uphill, you lose on the down hill.

I have real reservations about sampling fast enough with a PIC to catch both
slopes of a bump in the road, though, which could easily end up with your
device thinking the car is sinking into the earth or flying off into the
sky!

Mike H.

_________________________________________________________________
Help STOP SPAM with the new MSN 8 and get 2 months FREE*
http://join.msn.com/?page=features/junkmail

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email listservspamspam_OUTmitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@094234 by Steve Ruse

flavicon
face
In my experience the G-Tech Pro has pretty poor accuracy, I'm curious how much
better the "competition" model is. If it can hit 5/100 of a second for quarter
mile times, I would be impressed. Of course, judging horsepower this way is
very difficult. Its pretty tough to find a truly flat stip of pavement to test
on, and this device makes no attempt to account for wind & drag, etc. They are
pretty consistent though, which is good enough if you are tuning/modifying a
car.

In college we designed/built a PIC powered onboard computer that transmitted
data realtime back to a base station for our FSAE car. One of the devices we
were taking data from on the car was a two axis Analog Devices accelerometer.
We made no attempt at determining speed though...its much easier to just read a
pulse from the transmission. Vibration is an issue, we just used a bandpass
filter to get rid of the frequencies we didn't want.

Steve


> {Original Message removed}

2003\06\19@094448 by D. Jay Newman

flavicon
face
> Seems a gent made an accelerometer for his new Beetle...Two axes only,
> but the process could be expanded.

I like how he built it into a cup so it could be put into a cup-holder.  :)

> I don't know if you can do vector math with a PIC, but it seems to me

You can do vector math with an abacus. It takes a bit of thinking, but
it can easily be done on a PIC.

> I have real reservations about sampling fast enough with a PIC to catch both
> slopes of a bump in the road, though, which could easily end up with your
> device thinking the car is sinking into the earth or flying off into the
> sky!

I think that most of the MEMs accerometers sample only at around 400 Hz,
but I'm not sure.

I would use a couple of gyroscopes if I really wanted to do inertial
positioning.

I'm planning on using accelerometers to make sure my robot is going in
a straight line. Yes, a gyroscope would be better for that also, but
I *have* multiple accelerometers and don't have any gyroscopes...  :)

Hmmm. Combined with a compass it might make it good enough for simple
inertial guidance inside a house.
--
D. Jay Newman          !
@spam@jayKILLspamspamsprucegrove.com    ! Xander: Giles, don't make cave-slayer unhappy.
http://enerd.ws/~jay/  !

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email KILLspamlistservKILLspamspammitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@095312 by Duane

flavicon
face
Hello,

I just chimed in.  Where could one go about getting some of these
accelerometers and/or gyroscopes?

Thanks,
Duane


{Original Message removed}

2003\06\19@095727 by D. Jay Newman

flavicon
face
> I just chimed in.  Where could one go about getting some of these
> accelerometers and/or gyroscopes?

I got mine from Analog Systems as samples. So for my own peace of mind
I have to write up an article on an interesting use of one of these
chips.
--
D. Jay Newman          !
RemoveMEjayTakeThisOuTspamsprucegrove.com    ! Xander: Giles, don't make cave-slayer unhappy.
http://enerd.ws/~jay/  !

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email spamBeGonelistservspamBeGonespammitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@102225 by Daniel Serpell

flavicon
face
On Wed, Jun 18, 2003 at 10:34:51PM -0400, Marc Nicholas wrote:
>
> Bounce IR off the asphalt and then detect the scatter pattern. Vaguely
> unified signal == not moving. Combine with an accelerometer and you've got a
> good dead-reckoning system.

This can be done with an optical mouse. All the image processing is
already there, you only need to adjust the optics.

   Daniel.

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email TakeThisOuTlistservEraseMEspamspam_OUTmitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@104339 by Olin Lathrop

face picon face
> I don't know if you can do vector math with a PIC,

Yes, you can.  I've done it using 24 bit floating point on a 16F877.

> I have real reservations about sampling fast enough with a PIC to catch
> both slopes of a bump in the road, though, which could easily end up
> with your device thinking the car is sinking into the earth or flying
> off into the sky!

If you're making a straight line assumption, this isn't much of an issue
because you can low pass filter in hardware before sampling in the PIC.
If you are taking turning into account, you have to sample fast enough to
have only turned a small amount per sample.  Cars just don't turn that
fast, so even 10Hz sampling is probably fast enough.


*****************************************************************
Embed Inc, embedded system specialists in Littleton Massachusetts
(978) 742-9014, http://www.embedinc.com

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email RemoveMElistservspamTakeThisOuTmitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@104751 by Sean H. Breheny

face picon face
Hi all,

Since I'm also doing inertial navigation for our four-rotor helicopter
project which I recently mentioned in a few other threads, I thought I'd
chime in :-)

There are several problems in doing what you want to do. First is the
problem of how large the gravitational acceleration is on earth. If you do
not account absolutely perfectly for the direction of the gravity vector,
you will get error that accumulates very fast. For example, if you had an
accelerometer mounted horizontally in your car and then you tilted the car
1 degree (like going up a slight hill), you would now measure an
acceleration of 0.017 G when not accelerating. This translates to 0.16
meter/second^2, so if this persisted for 60 seconds, you would be about 10
m/s off in your velocity estimate and almost 300 meters off in your
position estimate.

The second problem is that the typical offsets on these accelerometers are
large and highly temperature dependent. Even if you measure the offset at
time zero and then subtract it off, a few degrees Celsius will be enough to
cause it to change by a few milliG.

There is also the problem that the output has some noise which has
significant low frequency components. This has the effect of also causing
an effective bias change over the time scale of seconds or tens of seconds.
(Also a problem if you plan on just sampling the bias and then subtracting
it out).

Inertial navigation works best when you have some absolute position sensor
(that need not have a high update rate) that you can use to update your
estimate. This is why GPS/INS combinations are now one of the standard ways
of doing navigation on high-end aircraft avionics systems. These work by
using GPS as the long term reference and INS for the short term. They also
have an error model for the INS and use the GPS to estimate its parameters
and constantly tune it.

Some of the ideas presented here (like combining this with gyros or using
two accelerometers to try to measure tilt, etc.) are good but any way you
approach this problem you are going to have to deal with several major
sources of error, most of which are highly temperature dependent. With
careful design you might be able to get a decent measurement of 1/4 mile
speed and time with cheap sensors, but it will be a challenging design task.

The idea of trying to determine when the car is no moving is good, but
remember, this will only give you an update when you stop. If you are
driving for 30 or 60 seconds, your error will be accumulating that whole time.

Sean

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email listservEraseMEspam.....mitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@112203 by Olin Lathrop

face picon face
> I would like to make a speedometer for a vehicle but would like to make
> one that needs no calibration (per say).
>
> As the vehicle will not be travelling for long times, I am thinking of
> using one of the ADXL digital accellerometers and try to calculate the
> speed from the acceleration obtained.
>
> Is this achievable and if so, does anyone have an idea or links to the
> math involved.

Short answer: No.

I've had some experience tracking human head motion with a combination of
accellerometers and gyros.  The results were "good enough" over the 1 to 2
second interval we cared about, but I did run into a number of issues with
the accellerometers.

One big issue with accellerometers is that gravity can't be distinguished
from accelleration relative to the ground.  In theory, if you knew
precicely the orientation of the accellerometer and it was well
calibrated, you could subtract out gravity.  In practise, both the
orientation and calibration are unachievable to allow more than a few
seconds of tracking.  Note that even a 5 degree tilt from horizontal will
result in an apparent accelleration of .85m/S/S.  After 2 seconds you're
already off by 1.7m/S, and note that the distance error grows with the
square of time.

Another problem was that the zero point on these things kept drifting.
Even after each end was calibrated to gravity, the zero point was often
off by a few percent g.  Worse yet, the zero was a function of the recent
accelleration.  You can take an accellerometer and hold it level on a
table, then zero the reading in software.  Now point it up so that it
reads 1g for a few seconds, then return it to the level position.  You
will probably see a few percent g reading now, even though it was
calibrated to zero just seconds before.  Lateral jolts (like a vehicle
going over bumps) can also confuse the accelerometers.

Another problem is that you need to track rotation unless the vehicle is
known to go in a straight line.  The full 3D solution takes 6 sensors and
considerable cycles to compute.  I originally started with a 16F877 taking
readings every 10mS, but this had to be changed to every 20mS because
otherwise all the 3D math floating point computations took longer than the
time between readings.  And, I took a few short cuts with the 3D math
based on the characteristics of what we were tracking.  (In the end, 50Hz
sampling turned out to be better than 100Hz sampling for other reasons
anyway).

There are inertial navigation systems out there that do solve all these
problems, but they don't rely on cheap MEMS accellerometers.  They also
cost many K$.


*****************************************************************
Embed Inc, embedded system specialists in Littleton Massachusetts
(978) 742-9014, http://www.embedinc.com

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email EraseMElistservspammitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@131928 by Gary Pepper

flavicon
face
You can get the vehicle's speed from the on-board diagnostic(OBD-2)port of
any 1996 (model year) or later vehicle.  You have to know the vehicle's
OBD-2 protocol (e.g. J1850 VPW, J1850 PWM, ISO9141-2, KWP2000 or CAN) and
what 'command' to issue.  You can get this data in real time, with 5-10
updates per second.

Gary

{Original Message removed}

2003\06\19@132556 by Dave Tweed

face
flavicon
face
"Dennis J. Murray" <RemoveMEdjmurrayEraseMEspamEraseMEBELLATLANTIC.NET> wrote:
> I'm probably missing something obvious here, so forgive my ignorance.  I've
> played some with the Analog Devices ADXL series accelerometers and been
> quite impressed with them.  It seems like your idea of deriving speed from
> acceleration could be accomplished with a hybrid solution.
>
> Why not use BOTH an accelerometer and a wheel/axle/driveshaft sensor??  Use
> the wheel sensor to detect "at rest" condition and very slow acceleration?
> You could determine "instantaneous" speed from the ADXL chip and
> mathematically back into how many counts that speed represents on the wheel
> sensor (i.e. recalibrate the wheel sensor?).  When you have constant speed
> OR very slow acceleration, use the recalibrated wheel sensor for that data.
> When the acceleration gets high enough to be in the linear (acceptable?)
> range of the ADXL, then switch over to that chip for your data.  You
> wouldn't have to recalibrate the wheel sensor every reading of the ADXL,
> just when the reading dropped below some arbitary acceleration threshold.
> None of this is difficult programming, so we're not talking rocket science
> here.

Now you're starting to get into the realm of Kalman filtering, a very
powerful technique for making use of noisy or incomplete information about
a system. Conceptually, it's very simple, but there are tons of details to
track down in order to create a working system.

The basic idea is that you set up a mathematical model of the system (the
vehicle). You use the sensor inputs, as many different ones as possible, to
update the state of the model -- but you also "weight" the values from the
various sensors based on the state of the model, giving greater weight to
those sensors that are known to be more accurate in that regime of
operation.

Furthermore, the weighting functions themselves can vary dynamically as the
filter tries to minimize the net error from all sensors. The filter can
learn about and compensate for things like sensor offset and gain errors.
For example, you could add a temperature sensor that would allow the filter
to predict temperature-related errors in the other sensors. It's all a
question of just how complete you want to make the model.

The outputs of the filter (acceleration, velocity, position, etc.) are the
state variables of the model, not the sensor inputs themselves.

-- Dave Tweed

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email RemoveMElistservspam_OUTspamKILLspammitvma.mit.edu with SET PICList DIGEST in the body

2003\06\19@135547 by Picdude

flavicon
face
On Thursday 19 June 2003 08:39, Steve Ruse scribbled:
> In my experience the G-Tech Pro has pretty poor accuracy,
FWIW, when these were first introduced, a lot of the car mags, car sites, & people tested them against the traditional 1/4-mile track timers.  Most found similar results of about 0.2s less than the track timer.  For <$100 US, I consider that quite accurate.  And it sure beats a buddy-with-a-stopwatch! :-)


> I'm curious how much
> better the "competition" model is. If it can hit 5/100 of a second for
> quarter mile times, I would be impressed. Of course, judging horsepower
> this way is very difficult. Its pretty tough to find a truly flat stip of
> pavement to test on, and this device makes no attempt to account for wind &
> drag, etc.

You can always derive this using any drag-timer software such as Desktop Dyno, etc.  But you're right ... the horsepower figures are pretty useless.


>  They are pretty consistent though, which is good enough if you
> are tuning/modifying a car.

Yes, and also if they're consistent, you can calibrate/correct for any inaccuracy.


> In college we designed/built a PIC powered onboard computer that
> transmitted data realtime back to a base station for our FSAE car. One of
> the devices we were taking data from on the car was a two axis Analog
> Devices accelerometer. We made no attempt at determining speed though...its
> much easier to just read a pulse from the transmission.

It is when your car doesn't also require a VSS for an EFI-ECU signal.  There's only one spot in the transmission.  They put out different signal types/levels.  I ended up building a little electronic converter on mine, but could not find a commercial solution for this otherwise.


> Vibration is an
> issue, we just used a bandpass filter to get rid of the frequencies we
> didn't want.
>
> Steve
>
> > {Original Message removed}

2003\06\19@152240 by Tal

flavicon
face
> FWIW, when these were first introduced, a lot of the car
> mags, car sites, &
> people tested them against the traditional 1/4-mile track
> timers.  Most found
> similar results of about 0.2s less than the track timer.

100$ for 0.2s reduction in 1/4-mile time seems like a great investment
for me. Imagine what 2000$ could do.

;-)

Tal

{Quote hidden}

> > > {Original Message removed}

2003\06\19@154109 by Dale Botkin

flavicon
face
On Thu, 19 Jun 2003, Tal wrote:

> > FWIW, when these were first introduced, a lot of the car > mags, car
> sites, & > people tested them against the traditional 1/4-mile track >
> timers.  Most found > similar results of about 0.2s less than the
> track timer.
>
> 100$ for 0.2s reduction in 1/4-mile time seems like a great investment
> for me. Imagine what 2000$ could do.

Pretty linear, I think.  I bet I could trim 4 seconds off my truck's
quarter mile time with $2K!  8-)  Of course that only applies to the FIRST
four second reduction, the next four would be substantially more expensive
I'm afraid.

Dale
--
It's a thankless job, but I've got a lot of Karma to burn off.
Get a PicoKeyer: http://www.hamgadgets.com

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email RemoveMElistservTakeThisOuTspamspammitvma.mit.edu with SET PICList DIGEST in the body

2003\06\20@060802 by Peter L. Peres

picon face
Whatever happened to doppler speedometers mounted under the car ? I
thought they would be pretty common by now ? I'd be tempted to use one of
those radar/doppler door opener modules (the kind that works with rf) and
just use a pic as a frequency counter on the output signal. Some
calculations are in order but the error of the oscillator should keep
total doppler error well below 1% and counter error at quartz accuracy.
For a quick test just wire headphones to the unit held outside the window
pointing down and backwards to hear the tone (probably not by the person
who is driving).

Peter

--
http://www.piclist.com hint: To leave the PICList
EraseMEpiclist-unsubscribe-requestspamspamspamBeGonemitvma.mit.edu>

2003\06\20@060812 by Peter L. Peres

picon face
>Hmmm. Combined with a compass it might make it good enough for simple
>inertial guidance inside a house.

A compass will show bs inside any house with metal objects in it (such as
rebars in the floor and ceiling etc and furniture objects with magnetic
metal in them). If you also have dc circuits in the house your north may
be *up* or such. Exceptions confirm the rule, beware.

Peter

--
http://www.piclist.com hint: To leave the PICList
RemoveMEpiclist-unsubscribe-requestKILLspamspammitvma.mit.edu>

2003\06\20@092201 by D. Jay Newman

flavicon
face
> >Hmmm. Combined with a compass it might make it good enough for simple
> >inertial guidance inside a house.
>
> A compass will show bs inside any house with metal objects in it (such as
> rebars in the floor and ceiling etc and furniture objects with magnetic
> metal in them). If you also have dc circuits in the house your north may
> be *up* or such. Exceptions confirm the rule, beware.

Ayup. But I believe that "combined" is the important word here.

If you can assume that the floor of the house is reasonably flat, then
the combination of the compass and accelerometers may work for a small
robot, especially if the robot has some means of localization and
"remembers" the magnetic parameters at the exceptional locations.

I think that I'd rather use an optical mouse to keep my robot headed
straight and the other sensors to determine landmarks (for example,
if the robot is near the parrot's cage, then the compass would go
wonky).

Now all I have to do is figure out how to use an optical mouse...
--
D. Jay Newman          !
jaySTOPspamspamspam_OUTsprucegrove.com    ! Xander: Giles, don't make cave-slayer unhappy.
http://enerd.ws/~jay/  !

--
http://www.piclist.com hint: To leave the PICList
spamBeGonepiclist-unsubscribe-requestSTOPspamspamEraseMEmitvma.mit.edu>

More... (looser matching)
- Last day of these posts
- In 2003 , 2004 only
- Today
- New search...