Back to Home

Extended Kalman Filter (EKF) for Sensor Fusion

Unlike standard textbook implementations, this EKF is engineered for production-grade numerical stability, specifically designed to prevent covariance divergence in high-dimensional state spaces and floating-point edge cases.

C++17Eigen3GTestCMake
GitHub

Key Engineering Features

  • Joseph Form Covariance Update: Utilizes the robust Joseph form for the measurement update step. This guarantees that the state covariance matrix P remains positive semi-definite and symmetric, even in the presence of severe floating-point rounding errors.
  • Cholesky Decomposition: Replaces direct matrix inversion with Eigen’s LLT (Cholesky) decomposition S.llt().solve() when computing the Kalman Gain K. This provides superior numerical precision and computational efficiency.
  • Modern C++ Architecture: Clean API utilizing std::function for injecting non-linear state transition and measurement functions.
  • Hermetic Build System: Employs CMake FetchContent to automatically manage dependencies (Eigen3, Google Test), ensuring deterministic, zero-friction builds across Linux, macOS, and Windows.