RRT_exploration

Implementation of autonomous exploration based on RRT (Rapidly Exploring Trees) algorithm using ROS2 and SLAM Toolbox by Nav2 framework. Algorithm is heavily inspired by the paper by Umari et al. published on IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) in 2017.

This project is just an addition to my Mobile Robot Programming experiments and shouldn't be used as a state-of-the-art exploration algorithm for the real system.

It was tested on Ubuntu 22.04, with ROS2 Humble and Gazebo Classic.

Warning: Gazebo Classic is outdated. This project may be moved to the Gazebo Fortress/Harmonic in the future.

Dependencies

This project relies on external software:

• nanoflann C++11 header-only library for building KD-Trees. The code is already contained in the project.
• PCL Cloud Library for frontier clustering. Refer to installation instructions here.

How to build

Run from the project root directory:

home~$ colcon build && source install/setup.bash

How to run

home~$ ros2 launch single_agent_rrt rrt_exploration.launch.py

Current issues

• Algorithm finds frontiers very fast in the beginning and as the robot progresses but slows down exponentially when there's not much left to explore.
• Rarely occuring bug in local_frontier_detector which I still need to investigate.
• Current implementation of RRT trees visualization works very slowly in rviz2.

Further steps in implementation

• Previous version of the code used nodes composition. I am planning on bringing it back for convenience.
• An extra frontier detection rate observation block shall be added which tracks the frontier update rate. When the map is almost fully explored and frontier detection rate becomes slow enough to trigger a threshold the algorithm switches from building RRT to bruteforce search.
• A special hash map tracking the already explored areas shall be added to filter out these areas in filter node.
• Currently I added a custom Costmap2D layer which fixes a bug currently existing in Gazebo Classic simulation. If the bug is fixed for you or you use other versions just disable the layer in nav2_params:

custom_layer:
        plugin: "nav2_custom_costmap_plugin::CustomLayer"
        enabled: False