GTSAM documentation

Blog

Snapshot of the GTSAM 4.3 landing page and release overview.

GTSAM 4.3 is available

Reported CUDA optimizer speedups on NVIDIA A100 benchmarks.

GTSAM Goes GPU: A CUDA Backend for Nonlinear Optimization

FAST-Sync initialization of an MIT pose graph.

FAST-Sync: A Fast Start for Group Synchronization on Any Matrix Lie Group

Ground-truth camera images compared with Gaussian-splat renderings.

3D Gaussian Splatting meets Factor Graph SLAM

Dense scene reconstruction from VGGT-SLAM.

Bringing Geometric Foundation Models to SLAM: VGGT-SLAM and SL(4) Factor Graph Optimization

RTK GNSS trajectories and position errors on Tokyo driving sequences.

Toward Centimeter-Level Positioning with GTSAM: Double-Difference Factors for RTK GNSS

Taxonomy of global optimization solvers for 3D vision.

Certifiable Optimization, Part 3: Global Solvers for 3D Vision

Certifiable factor graph optimization framework.

Certifiable Optimization, Part 2: Certifiable Factor Graph Optimization

Chordal decomposition framework for certifiable optimization.

Certifiable Optimization, Part 1: Exploiting Chordal Sparsity with Bayes Trees

Quadruped trajectory optimized with inequality constraints.

Speaking of Legged Robots

Continuous-time Gaussian-process trajectory and uncertainty.

Smoothing Out the Edges: Continuous-Time State Estimation Tools in GTSAM via Gaussian Processes

Geometry of a quadratically constrained quadratic program.

Quadratic Programs and QCQPs in GTSAM

Pose and landmark state used for equivariant visual-inertial odometry.

The Manifold Kalman Filter Hierarchy, Part 4: Awesome Equivariant Filters!

Group action on a manifold for equivariant filtering.

The Manifold Kalman Filter Hierarchy, Part 3: Equivariant Filters

Simultaneous trajectory estimation and planning using factor graphs.

Factor Graphs and "World Models"

Pose and landmark factor graph illustrating custom measurement factors.

CustomFactor in Python, C++, and MATLAB

Compressed Bayes-tree support for incremental covariance queries.

Incremental Fast Covariance with iSAM2

Bayes-tree structure used for localized covariance recovery.

Fast Covariance Recovery with Bayes Trees

Estimated body and foot trajectories during stair traversal.

The Manifold Kalman Filter Hierarchy, Part 2: Legged State Estimation

Prediction and update steps for Kalman filtering on a manifold.

The Manifold Kalman Filter Hierarchy, Part 1

Triangulation error comparison as the number of cameras increases.

LOST in Triangulation

Covariance propagation when composing uncertain poses.

Reducing the uncertainty about the uncertainties, part 3: Adjoints and covariances

Gaussian uncertainty represented on a manifold.

Reducing the uncertainty about the uncertainties, part 2: Frames and manifolds

Illustration of uncertainty in robot state estimation.

Reducing the uncertainty about the uncertainties, part 1: Linear and nonlinear

Mount Rainier reconstruction from relative height measurements.

Mount Rainier's Eigenvectors

Cayley-coordinate distortion compared with rotation angle.

Releasing GTSAM 4.0.3

Coordinate frames illustrating pose and frame conventions.

Geometry and Variable Naming Conventions

SLAM factor graph with robot poses and landmarks.

What are Factor Graphs?

Factor graph for linear quadratic regulator control.

LQR Control Using Factor Graphs

Image feature matches used to illustrate robust estimation.

Look Ma, No RANSAC

Cassie bipedal robot used for legged state estimation.

Legged Robot State Estimation Using Factor Graphs, Part I

Launching gtsam.org

Moving to Github!