Skip to content

Latest commit

 

History

History
71 lines (41 loc) · 2.9 KB

README.md

File metadata and controls

71 lines (41 loc) · 2.9 KB

turtlebot-detect-slam

This repository contains documentation and resources for doing visual SLAM (simultaneous localization and mapping) on a TurtleBot3 Waffle Pi and ROS. This repository is a work in progress and will be constantly changing as I research and experiment with different packages and tools.

Setup

Turtlebot Setup

  • Follow the setup guide in the ROBOTIS e-manual.

    • Everything in this repository is tested using Ubuntu 20.04 with ROS Noetic on the Remote PC, and the provided Raspbian disk image in the manual (Section 6.2.1.2).
    • The Remote PC you use should have a somewhat powerful Nvidia GPU if you want the segmentation/detection packages to run at a respectable framerate.

Object Detection Setup - YOLOv3

Optional Setup

  • I recommend installing ROS web video server so you can easily look at different image topics in your web browser. Otherwise you can use rqt_image_view or something.

sudo apt-get install ros-melodic-web-video-server

Usage

  • Start up ROS master. SSH into the turtlebot to start up its sensors and its camera with the following commands.

roslaunch turtlebot3_bringup turtlebot3_robot.launch

roslaunch raspicam_node camerav2_1280x960.launch

  • Run the launch file (on the Remote PC).

roslaunch visual_slam classify_and_locate.launch

  • To use custom weights, either edit yolo_v3.launch in the darknet_ros package or make a new launch file that points to the proper weights/configuration files. Make sure to edit the launch file in this package to point to the new darknet_ros launch files as well.

Simulation

This package also has support for Gazebo simulation, with the Turtlebot3 Simulations package as a prerequisite. The launch file opens an empty world by default.

roslaunch visual_slam gazebo_classify_and_locate.launch

Classes in weights files

bevo.weights

  • bevo

turtlebot.weights

  • turtlebot_burger

multiclass.weights

  • box
  • bevo
  • bottle
  • footwear
  • turtlebot

multiclass-v3.weights

Updated version of multiclass.weights (no footwear)

  • box
  • bevo
  • bottle
  • turtlebot