Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

this is my first post so hopefully this topic is considered OK.

Background: In class we were using a laser (mounted on a planar robot) to measure various profiles of a sample underneath. The system was already calibrated when I used it but I started wondering how this calibration would be performed (i.e. the math behind it). Here is what I came up with for a setup (simplified):

My Setup Idea

Setup: Assume the calibration gauge is a flat plate (blue, at angle $\alpha$ from x-axis and angle $\beta$ from y-axis, ideally it would be parallel to the x-y plane), then there will be: some initial height offset at the robot's origin ($\vec{Z_0}$, unknown), the laser's position attached to the planar robot ($\vec{P}$, known), the laser's measurement vector ($\vec{D_m}$, |$\vec{D_m}|$ known), the "true" height directly below the laser/robot ($\vec{D_t}$, unknown), and the error between the two ($\vec{E}$, unknown).

Ideally, the laser is to be mounted such that the output beam is parallel to the z-axis but I'm sure there are some mounting angle errors (error angle $\phi$ from the z-axis in x-z plane and angle $\psi$ from the z-axis in the y-z plane).

$i$ is the point where the laser makes contact with the plate (x,y,z).

The "known" variables are |$\vec{D_m}$|, P_x, and P_y.

The unknown variables I need to "calibrate" the system (so I can calculate $\vec{D_t}$) are $\vec{Z_0}$ and the angles $\alpha, \beta, \phi$, and $\psi$.

My Attempt to Figure It Out: First, I found where the laser would intersect the plate/plane based on the robot's (x,y) position: $i = ( P_x + \Phi\frac{AP_x + BP_y + Z_0}{1-A\Phi-B\Psi}, P_y + \Psi\frac{AP_x + BP_y + Z_0}{1-A\Phi-B\Psi}, \frac{AP_x + BP_y + Z_0}{1-A\Phi-B\Psi})$

where $A$ is the slope of the plane along the x-direction, $B$ is the slope of the plane along the y-direction, $\Phi$ is the slope of $\vec{D_m}$'s x-component along the z-direction, and $\Psi$ is the slope of $\vec{D_m}$'s y-component along the z-direction. Using these instead of the angles made the equations cleaner.

The measurement vector: $\vec{D_m} = (\Phi\frac{AP_x + BP_y + Z_0}{1-A\Phi-B\Psi},\Psi\frac{AP_x + BP_y + Z_0}{1-A\Phi-B\Psi}, \frac{AP_x + BP_y + Z_0}{1-A\Phi-B\Psi})$

The Calibration Equations (need 5)

  1. Take measurements along x-axis, regress to find slope, set equal to $\frac{\partial|\vec{D_m}|}{\partial P_x}$.

  2. Take measurements along y-axis, regress to find slope, set equal to $\frac{\partial|\vec{D_m}|}{\partial P_y}$.

  3. Move to origin ($P_x = P_y = 0$ to eliminates variables), take measurement set equal to |$\vec{D_m}$|.

  4. Have two points on the plane ($p_1, p_2$) with a known distance between them ($\Delta$). Move the robot so that the laser dot hits the first point, save ($P_{x1}, P_{y1}$), move the laser to the second point, save that position information. Use $P_{x1}, P_{y1}, P_{x2}, P_{y2}$ in the equation: $\Delta = |i(P_{x1}, P_{y1}) - i(P_{x2}, P_{y2})|$.

  5. ...? I'm stuck.

Final Thoughts I thought this was pretty interesting and not being able to solve it has been bugging me. :) I have no idea how the system was actually calibrated but this is how I set it up. If there is a far easier way or if someone knows how these types of systems are actually calibrated please let me know (it would be an interesting read) but I would also like to figure out that fifth equation using my way. I'm not a math major so I figured this would be the place to go!

Thanks!

share|improve this question

1 Answer 1

First, a remark: if $\vec Z_0$ is indeed an unknown vector, then you have $7$ unknown scalars in the problem: $(\vec Z_0)_x,(\vec Z_0)_y,(\vec Z_0)_z,\alpha,\beta,\phi,\psi$, which means you'd need $7$ scalar equations, not five.

But suppose that $\vec Z_0$ is perfectly vertical, so only $(\vec Z_0)_z$ is unknown. And suppose even that we somehow know $\alpha=\beta=\phi=0$. Even in this situation we cannot determine $(\vec Z_0)_z$ and $\psi$ from any measurements of $|\vec D_m|$. Indeed, no matter where we move the robot by changing $P_x$ and $P_y$, the measurement is always $|\vec D_m|=|(\vec Z_0)_z|/\cos \psi$. A robot with $|(\vec Z_0)_z|=2$, $\psi=0$ will always give the same measurement as a robot with $|(\vec Z_0)_z|=1$, $\psi=\pi/3$.

Perhaps your robot actually measures something in addition (or instead of) $|\vec D_m|$.

share|improve this answer
    
Wow, I hadn't even noticed someone had answered; thanks for the response. Yes, it could very well be measuring something else (all I saw was the profile data but I was curious on how the system was calibrated to correct for various angle misalignments). I will ask my professor to see if he has any idea and update my post with any new info. Thanks again. –  john Aug 8 '12 at 17:01

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.