 |
Welcome to the Uncertainty Team |
Our Mailing List: psas-uncertainty
We meet weekly on Wednesdays in Fab 155 at 7:25pm. (Also check the Schedule page)
Our Current Status
ToDo List
Questions/Answers
PSAS Rocket Model
Roadmap
The uncertainty team is tasked with achieving things that are unpredictable and very challenging. Of course, things which are certain and easily understood will be handled by other Teams.
We're doing research and implementation of probabilistic algorithms. In particular we are interested in all sorts of filtering: kalman unscented, kalman extended, sigma point kalman, particle and others. These are useful in estimating the position and orientation of a rocket given a diverse array of sensor inputs, all of which are incorrect in their own way. Without such estimation and eventually feeding back this information to our system, controlling a rocket's trajectory can be hugely difficult.
Please join our mailing list, or better yet stop by one of our meetings, we'd love to have your help!
Current Action Items (i.e. What we're currently working on)
- Research: read books/papers, find other resources, and share what you learn.
- Prototype: Running data from last flight through ReBEL toolkit, implementing extended kalman and unscented kalman in C or C++ code
Meeting Minutes
| 2006 | |
|---|---|
| Date | Summary |
| 8/02/06 | Judy, Tim and Markus talked about ekf, ukf within {{{ReBEL}}} and what the model should look like. Discussed issues regarding gains of gyros and misalignment. Tim wants to think about what the consequences of not including the gains in the error model. The gains typically drift slowly whereas bias' change quickly. Talked with Bart, he mentioned that he would like to do plain old kalman filtering first and then just move on to using one of the Bayesian Filtering Libraries, either BFL or Bayes++ to develop a true particle filter. We also had a new member join our group, Abraham. Welcome! |
| 7/26/06 | Judy & Tim discussed the next moves. Probably getting {{{ReBEL}}} work on simulated data. Also work on bias estimation using GPS data for INS integration. |
| 6/21/06 | Updated this wiki, and books/papers sections |
| 6/14/06 | Talked with Tim, Jamey about using Octave, will install it and try to load {{{ReBEL}}} within it |
| Markus commented that {{{ReBEL}}} would most probably need to get ported to Octave and to understand using it within Matlab first | |
| Judy brought more books on Kalman filtering and is updating the Papers and Books sections of this website as well as coming up to speed | |
Reading Lists
Source code libraries and implementations:
- Bayesian Filtering Library
- Bayes++ Bayesian Filter Classes
- 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
Local Resources
- Introduction to the Kalman fiter
- Introduction to state space representations
- Example:INS Aiding and Error Analysis in 1-D
Other Useful Information
Prior Work
- PriorWork
- ActiveGuidance system links and notes.
Attachments: