Use if for better SLAM, especially large for large scale like cars and drone

describe what can people use it for, or how it makes existing tasks easier/safer e.t.c (markdown supported)

The problem it solves

tell us about any specific bug or hurdle you ran into while building this project. how did you get over it? (markdown supported)

Challenges we ran into

FrostHack 2025

YAML

ROS2

VINS Fusion

Robot Localization

Mono VINS Fusion Based Localization

Visual-Inertial Odometry (VINS-Fusion), Adaptive Monte Carlo Localization (AMCL), and IMU-based filtering (EKF/UKF) for robust localization of TortoiseBot in ROS 2 Humble.

The localization system consists of the following components:

VINS-Fusion: Provides relative pose estimation using camera and IMU data.
AMCL: Corrects global localization by comparing laser scan data with a known map.
IMU + EKF/UKF: Fuses acceleration and gyroscope data for short-term stability.
Fusion Strategy: The robot_localization package (EKF/UKF) is used to merge multiple odometry sources.
Data Flow
[Camera + IMU] --(VINS-Fusion)--> /odometry
[Lidar] --(AMCL)--> /amcl_pose
[IMU] --> /imu_plugin/out
[Wheel Odometry] --> /odom
All Inputs --> (EKF/UKF) --> /filtered_odom

Autonomous Global  Localization for TortoiseBot

Humble

Mono VINS Fusion Based Localization

Using VINS Fusion for 1 mu and 1 camera to get pose and passing it along with imu and amcl_pose to and efk node to localize robot

Mono VINS Fusion Based Localization

Using VINS Fusion for 1 mu and 1 camera to get pose and passing it along with imu and amcl_pose to and efk node to localize robot

The problem Mono VINS Fusion Based Localization solves

Autonomous Global Localization for TortoiseBot

Discussion

Builders also viewed