MSCEqF v1.0

MSCEqF: Multi State Constraint Equivariant Filter

MSCEqF is a multi-state constraint equivariant filter for visual-inertial navigation.
It is based on the recent advances in equivaraint inertial navigation systems [1, 2, 3, 4].

Features

Design features

• Developed as a pure C++ library with ROS1 and ROS2 wrappers available

Filter features

• Supports online camera extrinsic and intrinsic parameters calibration
• Supports unit-plane projection method
• Supports anchored euclidean, anchored inverse depth and anchored polar feature representation methods
• Includes a static initialization routine as well as parametric initialization with custom origin
• Includes an equivariant zero velocity update routine
• Naturally consistent without the need for FEJ or OC

Vision frontend features

• OpenCV based
• Supports a grid-based multi-thread parallel feature extraction
• Supports different features detector including FAST and Shi-Tomasi
• Supports different image enhancment tecniques, including Histogram and CLAHE

Comparison to standard MSCKF on Euroc dataset

References

[1] van Goor, Pieter, Tarek Hamel, and Robert Mahony. "Equivariant filter (eqf)." IEEE Transactions on Automatic Control (2022).

[2] Fornasier, Alessandro, et al. "Equivariant filter design for inertial navigation systems with input measurement biases." 2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022.

[3] Fornasier, Alessandro, et al. "Overcoming Bias: Equivariant Filter Design for Biased Attitude Estimation with Online Calibration." IEEE Robotics and Automation Letters 7.4 (2022): 12118-12125.

[4] Fornasier, Alessandro, et al. "Equivariant Symmetries for Inertial Navigation Systems." ArXiv preprint.