online_estimators#

TAG-K: Tail-Averaged Greedy Kaczmarz for Online Inertial Parameter Estimation

online_estimators is a Python package implementing Kaczmarz-family iterative solvers and classical estimation methods (RLS, KF) for real-time inertial parameter estimation in robotic systems, with built-in quadrotor and double-pendulum dynamics.

Note

Accepted at ICRA 2026 — IEEE International Conference on Robotics and Automation.

Key Features#

  • 16 online estimation algorithms with a unified iterate(A, b) → x API

  • Quadrotor dynamics — 13-state quaternion model calibrated to Crazyflie 2.1

  • Double-pendulum dynamics — 4-state model for benchmarking

  • LQR controller — discrete-time with automatic linearisation

  • Noise models — AWGN, Ornstein–Uhlenbeck, random walk

  • 7 reference trajectories — hover, figure-8, circle, ellipse, helix, Lissajous, spiral

  • Optional C++ backends via pybind11 (1.5×–1.9× faster on laptop CPUs)

Installation#

git clone https://github.com/A2R-Lab/TAG-K.git
cd TAG-K
pip install -e .                         # pure-Python
pip install -e ".[cpp]"                  # with C++ accelerated backends

Quick Start#

import numpy as np
from online_estimators.estimators import TAGK

rng = np.random.default_rng(0)
n = 5
A = rng.standard_normal((50, n))
x_true = rng.standard_normal(n)
b = A @ x_true

est = TAGK(n)
for _ in range(500):
    x_hat = est.iterate(A, b)

print("Error:", np.linalg.norm(x_hat - x_true))

API Reference

Links

Indices and tables#