Rico贾若童的博客

Python - Python 3.14

Python 3.14 is shaping up to be the most ambitious release since type hints landed, with concurrency and ergonomics getting the spotlight. Free-threaded CPython Ships both the tradit...

Posted by Rico's Nerd Cluster on October 26, 2025

Computer Vision - 3D Gaussian Splatting

3D Gaussian Splatting (3DGS) renders scenes in real time by replacing expensive ray marching with a dense cloud of Gaussian blobs. Each blob is projected onto the image plane (a “splat”) and blende...

Posted by Rico's Nerd Cluster on August 1, 2025

Robotics - Model Predictive Control

Resources https://www.bilibili.com/video/BV1LSXKYEEW3/?spm_id_from=333.337.search-card.all.click&vd_source=ae0bfd67b026e62fbc37ca190dfd1839 https://blog.csdn.net/qq_37746927/article/detai...

Posted by Rico's Nerd Cluster on July 1, 2025

Fitness - Eat, Move, and Burn Smarter

Eat, Move, and Burn Smarter: A Simple Guide to Metabolism Most of us think “eat less, move more” is the whole story of weight loss and health. But under the hood, your body is running a sophistica...

Posted by Rico's Nerd Cluster on May 1, 2025

Robotics - Moré-Thuente Line Search Algorithmk

Introduction Introduced in the early 1990s, line search and trust-region strategies have become standard tools for improving the robustness of iterative optimization methods used in LiDAR odometry...

Posted by Rico's Nerd Cluster on April 17, 2025

Robotics - Gauss Newton (GN) and Levenberg Marquardt (LM) Optimizers

Newton's Method, GN, LM optimizers

Newton’s Method Solving An Equation To find an arbitrary equation’s root $f(x) = 0$, We start from an arbitrary point $x_0$ that’s hopefully close to the solution, $x_s$ The main idea of Ne...

Posted by Rico's Nerd Cluster on April 14, 2025

Robotics - [3D SLAM - 5] LIO Hands-On

Issues That I Encountered Synchronization THere are cases where you lose IMU for a bit. There could be a lag from the liadr to the imu as well. What’s strange is ROS publisher ...

Posted by Rico's Nerd Cluster on April 8, 2025

Robotics - [3D SLAM - 5] Loosely and Tightly Coupled Lidar Inertial Odometry

FastLIO

Loosely vs Tightly Coupled LIO In loosely coupled LIO, the IMU and GPS are first fed into error state kalman filter which outputs a global pose estimate T_G. T_G is fed into a lidar odometer (such...

Posted by Rico's Nerd Cluster on March 11, 2025

Robotics - [3D SLAM - 4] Iterative Extended Kalman Filter

Prediction: xk = xk_1 + F_k (x_k-1 - x_k) Observation: h(x_k) = h(x_k-1) + H_x (xk-xk-1) innovation = z - h(x_k) My understanding is in innovation, we do it iteratively. x = get_kf_pose kf.updat...

Posted by Rico's Nerd Cluster on March 8, 2025

Robotics - [3D SLAM - 3] Lidar Odometry

Direct Lidar Odometry, NDT, Incremental NDT, Indirect Lidar Odometry, LOAM

Direct Lidar Odometry When given two scans, a direct NDT method means no features are extracted. The two scans would directly go through NDT for finding their relative poses. The workflow is as fo...

Posted by Rico's Nerd Cluster on March 4, 2025

Rico's Nerd Cluster