gpfilter.py
Science Score: 57.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
✓CITATION.cff file
Found CITATION.cff file -
✓codemeta.json file
Found codemeta.json file -
✓.zenodo.json file
Found .zenodo.json file -
✓DOI references
Found 1 DOI reference(s) in README -
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (8.3%) to scientific vocabulary
Repository
Basic Info
- Host: GitHub
- Owner: adrianLepp
- License: mit
- Language: Python
- Default Branch: main
- Size: 579 KB
Statistics
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 3
- Releases: 0
Metadata Files
README.md
gpfilter.py
gpfilter implements the gaussian process (GP) as discrete time state space model for hybrid state estimation with the interacting multiple model (IMM).
The bayes filter (BF) implementations are based on
filterpy (Unscented Kalman Filter)
and
torchfilter (Particle Filter)
.
Both libraries implement also other filters, which should be able to used here, but are not testet yet.
The torchfilter library is augmented with an interacting multiple model particle filter (IMM-PF) which is based on the existing particle filter implementation and the IMM-PF algorithm presented in [1].
Gaussian process state space models models are derived from GpyTorch.
The implementation based on torchfilter is to be favored, since everythin is based on pytorch and the parameter optimization could be done for the whole filter module at once. At the moment optimization is only done for the single gp models, but in future this should be implemented since it makes much more sence to tune the gp predictions according to the filter performance.
Installation
bash
$ git clone https://github.com/adrianLepp/gpfilter.py
$ cd gpfilter.py
$ pip install -e .
Use
To see how to use the models, check the files in the submodule examples, where one implentation for torchfilter and filterpy is done for a nonlinear water tank system.
References
[1] A. Lepp und D. Weidemann, "Interacting-Multiple-Model Partikelfilter zur Fehleridentifikation", in: ASIM 2022, Jan. 2022, S. 187–194. doi: 10.11128/arep.20.a2024. url: a2024.arep.20 OA.pdf
Owner
- Login: adrianLepp
- Kind: user
- Repositories: 1
- Profile: https://github.com/adrianLepp
Citation (CITATION.cff)
# This CITATION.cff file was generated with cffinit.
# Visit https://bit.ly/cffinit to generate yours today!
cff-version: 1.2.0
title: 'Interacting Multiple Model Gaussian Process Bayes Filter '
message: >-
If you use this software, please cite it using the
metadata from this file.
type: software
authors:
- given-names: Adrian
family-names: Lepp
repository-code: 'https://github.com/adrianLepp/imm-gp-bf.py'
abstract: >-
Implementation of the gaussian process state space model
(gp-ssm) to use in filters (ukf, pf, imm,...) and an
interacting multiple model particle filter (imm-pf) based
on torchfilter
keywords:
- gaussian process
- particle filter
- state estimation
- interacting multiple model
- unscented kalman filter
- hybrid state estimation
- pytorch
- torchfilter
- gpytorch
license: MIT
GitHub Events
Total
- Watch event: 1
- Push event: 2
Last Year
- Watch event: 1
- Push event: 2
Dependencies
- GPy *
- fannypack *
- filterpy *
- gpytorch *
- matplotlib *
- numpy *
- overrides *
- pandas *
- pytest *
- scipy *
- torch *
- torchfilter @ git+ssh://git@github.com/stanford-iprl-lab/torchfilter.git