How sfwa/ukf works
Project type and purpose
• This is an open‑source C++ library (“UKF”) that implements multiple variants of the Unscented Kalman Filter (standard UKF, square‑root UKF, optimized SR‑UKF) for state and parameter estimation.
• Primary goal: deliver a header‑only, template‑based library with zero or minimal dynamic memory use, suitable for embedded systems.
File structure
• Top‐level:
• CMakeLists.txt + .github/workflows/cmake.yml → build system and CI pipeline
• /UKF → core library headers (StateVector, MeasurementVector, Core, SquareRootCore, Integrators, Types, Config)
• benchmark/ → benchmarking executables against different UKF implementations
• examples/ → sample applications in C++ and Python wrappers (AHRS, SFWA‑UKF)
• test/ → unit test suite exercising core classes and algorithms
• patch/ → third‑party patch for Eigen
• Key architectural pattern: template‐metaprogrammed, policy‑based design. Core algorithms are generic on StateVector and MeasurementVector types and on Integrator policies.
README for additional insights
• Library depends on Eigen for linear algebra.
• Three extension points required by the user: StateVector specialization (fields), MeasurementVector specialization (fixed/dynamic), Core specialization (with choice of integrator).
• Build targets: library (header + inline code), unittest, benchmark, examples.