[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## [Help-gsl] How to Fit an intracable formula

**From**: |
Alex Brussee |

**Subject**: |
[Help-gsl] How to Fit an intracable formula |

**Date**: |
Sat, 03 Jul 2004 15:50:20 +0000 |

Hi All,

`I have a problem which i cannot solve. I used the GSL library (BFGS) to
``generate a Minimum Jerk Optimal solution in the form of a fifth order bezier
``curve going from a starting point to a via point to an end point. This
``generates a Data Set (x,y positions) on which i would like to fit a
``differential steering path.
`

`The differential steering path is generated using a formula that describes
``left and right wheel speeds seperately. This causes the Path generated to be
``intracable. Thus so are the partial derivatives of the distance (x,y)
``travelled.
`

`BFGS can estimate the partial derivatives, but in the gsl-library it seems
``that the partial derivatives must be supplied by the user (mf.df = &func).
``This works great in the case of the tracable polynomials, but how can i get
``the BFGS optimizer to work without these derivatives. (Or can someone give
``me the integration of my formula, beceause mathematica cannot)
`
The speed functions attached to the differential steering:
x(t)=Amp1*((1/(tau1-tau2))*exp(t/tau1)-(1/(tau1-tau2))*exp(t/tau2))
y(t)=Amp2*((1/(tau3-tau4))*exp(t/tau3)-(1/(tau3-tau4))*exp(t/tau4))

`(just to make clear that the resulting formula is indeed intracable, see
``also http://rossum.sourceforge.net/papers/DiffSteer/DiffSteer.html )
`
My question:

`1. Is there another (better) way in gsl to fit the generated path using
``Amp1,tau1,tau2,Amp2,tau3,tau4 to the data set ?
``2. Is there a way to obtain the partial derivatives by anther procedure, or
``did i overlook something ?
`
thnx for helping me out
Alex Brussee
:)
_________________________________________________________________
MSN Search, for accurate results! http://search.msn.nl

**[Help-gsl] How to Fit an intracable formula**,
*Alex Brussee* **<=**