GTSAM

GTSAM, the Georgia Tech Smoothing and Mapping library, is an open-source C++ library for solving large-scale optimization problems posed as factor graphs, with a focus on SLAM and robotics perception. It represents estimation problems as graphs where variables are nodes and measurements are factors, enabling probabilistic inference through nonlinear least-squares on manifolds.

The library provides both batch and incremental solvers. It supports nonlinear optimization on manifolds using Gauss-Newton and Levenberg-Marquardt methods, and features iSAM2 as the primary incremental solver for fast updates in streaming data scenarios. A rich collection of factor types is available, including priors, between factors, odometry, range-bearing, and IMU-related factors, along with various variable types such as Pose2, Pose3, Point2, and Point3. GTSAM leverages Lie group geometry and Bayes trees to achieve computational efficiency and modularity. It offers robust loss functions, multiple linear algebra backends, and interfaces to C++, Python, and MATLAB through wrappers. The library is built on the Eigen library for linear algebra.

History and licensing: GTSAM was developed at Georgia Tech’s robotics research community, led by Frank Dellaert, and has evolved through ongoing contributions from academia and industry. It is released as open-source software under a permissive license, enabling researchers and developers to adapt and extend it for SLAM, sensor fusion, and general estimation problems. GTSAM is widely used for both academic research and practical robotics applications.