Home

AwesomePKF

AwesomePKF is a Python library implementing the Pairwise Kalman Filter (PKF) and its nonlinear extensions for state estimation in pairwise Markov models.

The PKF generalises the classical Kalman filter by relaxing the conditional independence between states and observations, naturally accommodating phenomena such as measurement back-action and recursive observation dependencies — without resorting to cumbersome state-augmentation techniques.

---

Filters

Filter	Description
PKF	Pairwise Kalman Filter — exact, optimal (MMSE), fully recursive
EPKF	Extended PKF — first-order Taylor linearisation of the pairwise transition
UPKF	Unscented PKF — sigma-point propagation adapted to the pairwise structure
PPF	Pairwise Particle Filter — sequential Monte Carlo for pairwise models
UKF	Standard Unscented Kalman Filter (for comparison)
PF	Standard Particle Filter (for comparison)

---

Smoothers

Smoother	Description
Linear PKS	Linear pairwise Kalman smoother for linear-Gaussian models — six equivalent backward passes (RTS, BF, MBF, MF, DWY, VAR) behind one façade Linear_PKS(param, method=...)

See the Smoothers page for details.

---

Quick links

• 📦 Installation
• 🔢 Available Models
• 🔧 Filters & CLI
• 🔁 Smoothers
• 📓 Tutorials
• 🧠 Data-Driven Dynamics (NNModel)
• 📄 Paper Reproducibility
• 📚 API Reference

---

Installation

pip install awesomepkf

See the Installation page for all options.

---

Source & publication

GitHub	https://github.com/SDerrode/awesomepkf
PyPI	https://pypi.org/project/awesomepkf
Python	≥ 3.10
License	GNU AGPL v3

---

Reference paper

C. Fernandes, Z. Bouyahia, S. Derrode — "Non-linear extensions to the Gaussian Pairwise Kalman Filter", submitted to IEEE Transactions on Signal Processing, 2026.