Mailing List: psas-uncertainty
The uncertainty team is dedicated to doing stuff that we don't understand at all. We'll leave the stuff that we think we sort of understand to the other Teams.
Put another way, we're doing research and implementation of probabilistic algorithms. These are useful, for instance, to estimate the position and orientation of the rocket given a diverse array of sensor inputs, all of which are wrong in different ways. Without such estimation, controlling the rocket's trajectory is really, really, really hard.
Current to do list
- Research: read books, find other resources, and share what you learn.
Research plan
We want to have a process that updates estimates for the error model as we propagate the filter.
- Fix the GAINS prototype to rotate its orientation estimate through the angle between: the vector from last position estimate to current GPS coordinate; the vector from last position estimate to current INS position.
- Given a sequence of GPS height and Z-axis accelerometer measurements (and no rotation or horizontal acceleration), compute position estimate, bias, and gain.
- Extend simple filter to multiple, more complicated sensors.
- Include orientation of IMU platform in error model.
Projects
Data Fusion
Ok, maybe "Data Fusion" is dumb. But Redundancy, Inconsistency, Error, and Noise. From cacophony, we seek the truth. Here is just a little beginning.
In our current imu we make some redundant acceleration measurements. We want to use the extra data to refine our estimates and increase sensor reliability. Here is a first cut at integrating the x, y, and "q" axis accelerometer data.
- xyq.pdf (14K): Best estimates for X and Y given x, y, and q
GPS-aided Inertial Navigation System (GAINS)
See elsewhere for information on Inertial Navigation.
On 07 Oct 2003, in a joint meeting, we decided that implementation of GAINS has the same priority as switching to a hard-RealTime operating system and using our own ?open source GPS receiver.
On 06 Oct 2004, ?JameySharp announced a prototype GAINS implementation.
?JameySharp proposed to build a more complete implementation as a Google Summer of Code project, but his other proposal was accepted instead.
Resources
- Our effort at defining a project-wide coordinate system for navigation
- Some mathematical notation we use
- Our attempt to explain the Kalman filter
On 06 Oct 2004, BartMassey wrote about Bayesian Particle Filtering:
Haskell Resources
- Matt's DSP library: Modules for matrix manpulation, digital signal processing, spectral stimation and frequency estimation
- HAT: The Haskell Tracer: Source level tracer for ghc and nhc98
Reading List
Papers
- Study on GPS attitude determination system aided INS using adaptive Kalman filter: Hongwei Bian et al 2005 Meas. Sci. Technol. 16 2072-2079 doi:10.1088/0957-0233/16/10/024
- A Java Tool for Exploring State Estimation using the Kalman Filter: Declan Delaney and Tomas Ward. ISSC 2004, Belfast
Internet Resources
General introductory material:
- Wikipedia article
- Engineers Look to Kalman Filtering for Guidance: Barry Cipra, SIAM News, Vol. 26, No. 5, August 1993
- Some tutorials, references, and research on the Kalman filter at the Department of Computer Science at the University of North Carolina at Chapel Hill
- Kalman Filters at Connexions
- Taygeta's Kalman Filter Information and reading list
- Kalman Filtering, Dan Simon, Innovatia Software
Kalman filter extensions:
Implementations:
- ReBEL: Recursive Bayesian Filtering, Matlab toolkit, written by Rudolph van der Merwe and Eric A. Wan.
- Kalman filter toolbox for Matlab, written by Kevin Murphy
- The Kalmtool Toolbox Version 2 - for use with Matlab]
- Kalman filter for image sequence processing
Books
- Jay Farrell, Matthew Barth. The global positioning system and inertial navigation. New York : McGraw-Hill, c1999. ISBN 007022045X.
- Simon Haykin, Kalman Filtering and Neural Networks. Wiley, October 2001. ISBN: 0-471-36998-5
Meeting Minutes
| 2004 | |
|---|---|
| Date | Summary |
| 2004-10-21 | Investigating GAINS stuff |