DeKA: The Deterministic Kaczmarz Algorithm with Greedy Selection and Smoothing for Online Inertial Parameter Estimation

Overview

This project implements an online parameter estimation algorithm leveraging the Kaczmarz method to address overdetermined system identification problems efficiently. Our algorithm is compared with traditional parameter estimation methods like RLS and Kalman Filter.

Key Features

• Kaczmarz Method: A lightweight iterative method that uses incoming data points to continuously refine parameter estimates.
• Online Adaptability: The method operates in real-time, updating estimates as new data becomes available.
• Overdetermined Systems: Efficiently handles systems with more equations than unknowns, leveraging redundancy for improved accuracy.

Potential Impact

• Improved system control through real-time parameter updates.
• Reduced computational overhead compared to traditional recursive/batch methods.

Status

• Current Stage: Code cleanup.
• Next Steps: C++ implementation and testing in hardware.