Unified autonomous navigation framework

The repository implements a unified autonomous navigation framework for Turtlebot3 robot which includes perception, SLAM, and planning fully integrated. This was part of the final project of the course Autonomous Robotics (MT-0013) at UTEC@Peru. It is capable of handling dynamic env due to its two-step map generation from ORB-SLAM 2 and YOLOv8 which contributes to the dynamic local map only composed of people to meant to be avoided. Also, it includes two planers: RRT (global planner) and DWA (local planer) where the last one handles the dynamic obstacles.

Overview of components

Dependencies

• Octomap
• ros-navigation
• Ultralytics
• Depthimage_to_laserscan
• Turtlebot3, simulations, and messages

Steps for execution

1. Launch your selected Gazebo env and ORB-SLAM 2 for static mapping. Consider modifying the second launch according to your camera parameters.

   roslaunch your_gazebo_worlds_package your_gazebo_env.launch
   roslaunch orb_slam2_ros orb_slam2_rgbd_mapping.launch
   rosrun turtlebo3_teleop keyboard_teleop

2. Once your env has been fully explored, generate its occupancy grid and save both the grid file (.pgm) and ORB-SLAM 2 map (.bin).

   roslaunch orb_slam2_ros pointcloud_octomap.launch
   rosrun orb_slam2_ros filter_raw_map.py
   rosrun map_server map_saver --occ 90 --free 10 -f mymap map:=/filtered_raw
   rosservice call /orb_slam2_rgbd/save_map map.bin 

3. Load both maps, execute the planners, initiate YOLOv8 detection, and generate a virtual LiDAR scan from the depth image

   roslaunch orb_slam2_ros orb-slam2_rgbd_localization.launch
   rosrun orb_slam2_ros sendOdom2Map.py
   rosrun orb_slam2_ros filter_depth_people.py
   cd ros_motion_planning/src/sim_env/scripts/
   chmod u+x ./main.sh
   ./main.sh

Contact

*[email protected] *[email protected] *[email protected] *[email protected]