Can I use accelerometer to get position?

Started by September 26, 2006
I'm a new user of basicx
currently i have a biaxial accelerometer, can i use it to get the
position of a robot?
Accelerometer's are used to detect vibration. As an example you could use it
to tell if one wheel tilts too high on one side or the other.
-----Original Message-----
From: b... [mailto:b...]On Behalf Of
hankybanky123
Sent: Monday, September 25, 2006 8:26 PM
To: b...
Subject: [BasicX] Can I use accelerometer to get position?
I'm a new user of basicx
currently i have a biaxial accelerometer, can i use it to get the
position of a robot?
--- hankybanky123 wrote:

> I'm a new user of basicx
> currently i have a biaxial accelerometer, can i use
> it to get the
> position of a robot?
>
> good mornin
ya surely u can use that but confim first whether it
is workin properly or not? if any qurries then do
>

__________________________________________________________
Yahoo! India Answers: Share what you know. Learn something new
You can use two accelerometers to determine attitude related to gravity
(i.e. where is Down?), but not where the accelerometer is related to its
environment (where am I?) except as part of a Dead Reckoning algorithm.

http://www.memsic.com/memsic/pdfs/an-00mx-001.pdf
http://www.analog.com/en/cList/0,2880,764%255F%255F43,00.html
http://www.xbow.com/Support/appnotes.htm
Tom
--- In b..., "hankybanky123" wrote:
>
> I'm a new user of basicx
> currently i have a biaxial accelerometer, can i use it to get the
> position of a robot?
>

Yes, if you integrate acceleration once you will get velocity (V). If
you integrate the measured acceleration (A) signal twice, you can get
position (P). Be sure that the robot is at rest when you start
integration though because it needs known initial conditions of
position and velocity (which in this case is P(0)=0 and V(0)=0). You
will need to sample the acceleration at a known small time interval
(DT. i.e. DT=0.01 sec).

Here is a reasonable way to compute the first and second integrals to
yield velocity and position respectively at each time sample, k:

V(k)=V(k-1)+A(k)*DT
P(k)=P(k-1)+1/2*A(k)*DT*DT

These should look familiar to you from your physics text book. You
need to compute these equations for each axis to get Px and Py from

An important thing to note about getting position from an
accelerometer is that the error in position "integrates" What this
means is that if the noise or error in your accelerometer follows a
normal distribution (overestimates and underestimates equally) then
your position estimate should be reasonable. If however, the
accelerometer is biased (tends to overestimate more than
underestimate or vice versa) then the error in your position estimate
will grow exponentially. Any error is kept in your calculation
thtough the iteritive integration, so calculating position the
accelerometer can have large errors. But it still is a valid way to
get position. Give it a try!

Sami.
Double integration - drift - much too great. Position will be quite
inaccurate very quickly.

_____

From: b... [mailto:b...] On Behalf Of
Sam
Sent: Wednesday, September 27, 2006 12:01 PM
To: b...
Subject: [BasicX] Re: Can I use accelerometer to get position?

--- In basicx@yahoogroups. com,
"hankybanky123" wrote:
>
> I'm a new user of basicx
> currently i have a biaxial accelerometer, can i use it to get the
> position of a robot?
>

Yes, if you integrate acceleration once you will get velocity (V). If
you integrate the measured acceleration (A) signal twice, you can get
position (P). Be sure that the robot is at rest when you start
integration though because it needs known initial conditions of
position and velocity (which in this case is P(0)=0 and V(0)=0). You
will need to sample the acceleration at a known small time interval
(DT. i.e. DT=0.01 sec).

Here is a reasonable way to compute the first and second integrals to
yield velocity and position respectively at each time sample, k:

V(k)=V(k-1)+A(k)*DT
P(k)=P(k-1)+1/2*A(k)*DT*DT

These should look familiar to you from your physics text book. You
need to compute these equations for each axis to get Px and Py from

An important thing to note about getting position from an
accelerometer is that the error in position "integrates" What this
means is that if the noise or error in your accelerometer follows a
normal distribution (overestimates and underestimates equally) then
your position estimate should be reasonable. If however, the
accelerometer is biased (tends to overestimate more than
underestimate or vice versa) then the error in your position estimate
will grow exponentially. Any error is kept in your calculation
thtough the iteritive integration, so calculating position the
accelerometer can have large errors. But it still is a valid way to
get position. Give it a try!

Sami.