How World Models Can Help Robots Plan
Seven ways to bring world models into robot planning, from representation learning to closed-loop simulation.
Notes on robotics, computer vision, and world models.
Seven ways to bring world models into robot planning, from representation learning to closed-loop simulation.
A practical look at point cloud registration: correspondences, pose estimation, and outlier rejection.
How rotation conventions and left and right multiplication relate to coordinate frames and Euler angles.
Derivation, implementation details, and ablations for KLT optical flow.
A hands-on derivation of PnP with DLT, RANSAC, and Gauss–Newton, with code and experiments.
How edge representation, nearest-neighbor association, and point-to-tangent residuals fit into visual odometry.
A 2022 guide to open-source visual-inertial odometry projects, organized by their main contributions.
Six traditional clustering methods for segmenting objects in LiDAR point clouds.
A survey of point-cloud ground segmentation and four open-source approaches.
Ground-plane constraints and Jacobians for visual-inertial-wheel SLAM.
A derivation of midpoint and closed-form IMU–wheel odometry integration on SE(3).
A paper list and code collections for deep-learning-based 3D registration.
Calibrate a stereo camera in MATLAB and load the parameters in OpenCV.
Convert KITTI camera-frame odometry poses into the LiDAR frame.