Acknowledgement

This work was done collaborating with Or Tslil and Tal Feiner, please follow their works.

sparse_slam

ROS implementation of online semantic SLAM. Used for large unkown indoor environments. Object detector node is based on yolo and forked from https://github.com/zzh8829/yolov3-tf2 Line extration node is forked from https://github.com/bguplp/laser_line_extraction.git

Video example

The world:

Dependencies

• Python version > 3.5
• TensorFlow 2.* (GPU version)
• numpy
• pandas
• openCV
• pip3 install
• hector_mapping http://wiki.ros.org/hector_mapping
• yolov3-tf2 https://github.com/zzh8829/yolov3-tf2
• laser_line_extraction https://github.com/bguplp/laser_line_extraction.git
• depthimage_to_laserscan http://www.ros.org/wiki/depthimage_to_laserscan
• gazebo_models_worlds_collection https://github.com/chaolmu/gazebo_models_worlds_collection.git
• robot_description https://github.com/TalFeiner/robot_description

Setup

1. Copy, install and make the repository:

$ cd ~/your_catkin_worspace/src
$ git clone https://github.com/elbazam/sparse_slam.git
$ cd sparse_slam
$ pip3 install -e ./object_detector
$ pip3 install -e ./exits_detector
$ pip3 install -e ./room_slam
$ cd ..
$ catkin_make

2. Execute .sh file:

$ cd ~/your_catkin_worspace/src
$ https://github.com/TalFeiner/robot_description
$ cd ..
$ catkin_make
$ cd src/robot_description
$ chmod +x ./python3_alongside_python2.sh
$ ./python3_alongside_python2.sh

3. Replace the following contents:

• In <your_workspace_name>/src/sparse_slam/object_detector/object_detector/ros/object_localizer_node.py , line 89, replace the written path to your path:

• self.real_path = '/home/<Computer_name/<your_workspace_name>/src/yolov3-tf2>'

• Go to the xarco file ~/robot_description/blattoidea/urdf/kinect.xarco

• Change the orientation of the kinect to 0 0 0 from 0 0.3 0

Execution

Comparison to hector slam algorithm

Create files:

$ cd
$ mkdir '/home/melodic/data'
$ mkdir '/home/melodic/data/complex_sslam_data'
$ mkdir '/home/melodic/data/complex_hector_data'
$ mkdir '/home/melodic/data/complex_sslam_dt'
$ mkdir '/home/melodic/data/complex_hector_dt'

• Big environment:

roslaunch agent big_test.launch

• Small environment (the example above):

roslaunch agent Test.launch

Exectution of only room slam:

$ rosrun room_slam sslam_node.py

Pipelines

• Each package has its own Pipeline readme