https://github.com/fatlipp/toyslam

SLAM implementation from scratch w/o external graph optimization libs
https://github.com/fatlipp/toyslam

cuda gpu lidar-slam mapping odometry robotics slam

Last synced: 9 days ago
JSON representation

SLAM implementation from scratch w/o external graph optimization libs

• Host: GitHub
• URL: https://github.com/fatlipp/toyslam
• Owner: fatlipp
• Created: 2024-07-28T15:30:27.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-09-07T14:43:36.000Z (5 months ago)
• Last Synced: 2025-02-05T05:44:17.834Z (9 days ago)
• Topics: cuda, gpu, lidar-slam, mapping, odometry, robotics, slam
• Language: C++
• Homepage: https://medium.com/@fatlip/graph-based-slam-basics-f84501525f24
• Size: 250 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
## Simple SLAM Implementation

This project is a basic implementation of SLAM using a simulated 2D LiDAR. The goal is to demonstrate the fundamental principles of SLAM, including robot pose estimation and optimization using graph.

![SLAM](/assets/SLAM.png)

### Features

- 2D LiDAR Simulation: A virtual LiDAR sensor is simulated, providing distance measurements to surrounding objects. This data is used to map the environment and estimate the robot's pose.

 - Robot Pose Optimization: The robot's pose (position and orientation) is optimized using noisy landmarks and noisy robot positions.

 - Remote CPP optimizer:

   - 1. Build `cpp` folder:

     Example of using conan:

        * conan install . --output-folder=./build -s compiler.cppstd=gnu20 -s compiler.version=13 --build=missing

        * cmake --preset conan-release -DWITH_CUDA=OFF [ON/OFF]

        * cmake --build --preset conan-release

   - 2. Run: graph_optimizer HOST PORT ITERATIONS PIPELINE SOLVER (./bin/graph_optimizer "127.0.0.1" "8888" "50" cpu eigen)

        - ITERATIONS - [int value >= 1]

        - PIPELINE: [cpu/gpu]

        - SOLVER: [cuda/eigen] (if `PIPELINE == GPU` => `SOLVER = CUDA`)

   - 3. Run `python3 python/slam_main.py`

### Requirements

 - numpy: for matrix operations

 - CPP (optional, for remote optimization):

    * Eigen - for matrix operations

    * Conan 2 (optional) - packet manager

    * CUDA (optional)

### Pipeline

There is 4 different pipelines:

 - Python UI + Python optimizer

 - Python UI + CPP optimizer

 - Python UI + CPP optimizer (CUDA matrix solver)

 - Python UI + Full CUDA optimizer

### Key concepts

 - State Representation: The robot's state is represented by 3x3 matrix:

**[R t]**

**[0 1]**

 - Odometry:

**[R t]**

**[0 1]**

 - Graph2d: consists of robot poses in 2d and relative landmarks observations

 - OptGraph: consists of robot poses in 2d and relative landmarks observations, that used for optimization

 - 2 types of edges: [ODOM, LM]

### Simplification

Each point in the environment has its own ID, which is used to match landmarks

### Further development

 - Adding More Edge and Vertex types (2d, 3d, BA, Virtual Meas.)

 - Using a map for navigation

 - Performance optimization

 - CUDA opt improvements