Semantic Priority Navigation for Energy-Aware Mining Robots

This repository contains the code, configurations, and resources supporting the research paper "SemanticPriority Navigation for EnergyAware Mining Robots in ROS" (to be published). It implements a prioritized semantic navigation framework for Husky robots in mining environments using ROS and Gazebo simulations, focusing on perception (CNN detection, LiDAR segmentation, semantic fusion) and navigation (costmaps, local planners, experiment execution).

Prerequisites: Install ROS Noetic Follow the official installation guide: ROS Noetic Installation on Ubuntu.

System Requirements: Ubuntu 20.04 (recommended for ROS Noetic) Python 3.x with required libraries (see requirements.txt) ROS Noetic desktop-full installation

Setup Instructions:

1. Create a Catkin Workspace 1.1. Create the Workspace Directory mkdir -p ~/catkin_ws/src cd ~/catkin_ws/src 1.2. Initialize the Catkin Workspace catkin_init_workspace 1.3. Build the Workspace cd ~/catkin_ws catkin_make 1.4. Source the Workspace source ~/catkin_ws/devel/setup.bash

2. Install the Simulation Package sudo apt-get install ros-noetic-husky-simulator

3. Download Clearpath Gazebo Worlds: Clone the repository into your ROS workspace: cd ~/catkin_ws/src git clone https://github.com/clearpathrobotics/cpr_gazebo.git

4. Export Sensor Parameters Enable 3D LIDAR (Velodyne VLP-16) and RealSense camera: # 1. LIDAR 3D (Velodyne VLP-16) export HUSKY_LASER_3D_ENABLED=1 export HUSKY_LASER_3D_XYZ="0.0 0.0 0.0" export HUSKY_LASER_3D_RPY="0.0 0.0 0.0" # 2. CAMERA RealSense export HUSKY_URDF_EXTRAS=$HOME/Desktop/realsense.urdf.xacro

5. Copy Project Directories Copy the husky_navigation and husky_perception directories to your workspace: cp -r husky_navigation husky_perception ~/catkin_ws/src/ cd ~/catkin_ws catkin_make

6. Launch the Simulation Environment Deploy the robot in the Gazebo world: roslaunch cpr_inspection_gazebo inspection_world.launch platform:=husky

7. Launch the CNN Detection Node rosrun husky_perception cnn_detector.py

8. Launch the Obstacle Segmentation Node rosrun husky_perception obstacle_segmentation.py _visualize:=false

9. Launch the Navigation and Fusion Nodes roslaunch husky_navigation navigation.launch

10. Adjust Parameters Live Install and run the reconfigure tool: sudo apt-get install ros-noetic-rqt-reconfigure rosrun rqt_reconfigure rqt_reconfigure - From the GUI, select /semantic_fusion_node. - Adjust semantic_fusion_node parameters (e.g., thresholds, rays per priority) without restarting the node.

Running Experiments Once the system is running, execute experiments using the experiment_runner.py script with the following format: rosrun husky_navigation experiment_runner.py [ON/OFF] [obstacle_name] [test_number] - ON: Enables the perception system. - OFF: Disables the perception system. - obstacle_name: Name of the obstacle to avoid (e.g., person, backpack, chair). - test_number: Test number (e.g., 01, 05).

Example: rosrun husky_navigation experiment_runner.py OFF person 05 This runs test 5 with the perception system off, avoiding a person obstacle. Each experiment generates a .bag file in ~/experiment_bags for later analysis.

Analyzing Data Extract Data from Rosbags Run the analysis script to process .bag files: python3 analyze_experiments.py ~/experiment_bags --salida ~/resultados_analisis This generates an Excel file (experimental_results.xlsx) in the specified output directory for further analysis.

Generate Graphs Run the script to create summary figures (e.g., Figures 10, 11, 12 from the paper): python3 generate_graphs_results.py Output figures are saved in the figures folder.

Run Statistical Analysis Execute the following scripts to generate tables from the paper:

For Table 2 (assumption checks): python3 checking_assumptions_anova.py

For Table 4 (ANOVA results): python3 anova.py

For Tables 5 and 6 (Tukey HSD results): python3 Tukey.py

Repository Structure /husky_perception/cfg/: Dynamic reconfiguration files (e.g., SemanticFusionConfig.cfg.py). /husky_perception/scripts/: Perception nodes (e.g., cnn_detector.py, obstacle_segmentation.py, lidar_cnn_semantic_integration.py). /husky_navigation/config/: Navigation configuration files (e.g., costmap_common_params.yaml, global_costmap_params.yaml, teb_local_planner_params.yaml). /husky_navigation/launch/: Launch files (e.g., navigation.launch). /husky_navigation/scripts/: Navigation support scripts (e.g., clear_costmap_node.py, experiment_runner.py). /analysis/: Data analysis scripts (e.g., analyze_experiments.py, anova.py, checking_assumptions_anova.py, generate_graphs_results.py, Tukey.py). /figures/: Generated plots (e.g., TotalEffort_summary_barplot.png, trajectory_run_*.png). /data/: Experimental data (e.g., experimental_results.xlsx). /: Root files (this README.md, realsense.urdf.xacro).

Data Availability The synthetic data presented in this study are available on FigShare[](https://doi.org/10.6084/m9.figshare.29852780) in CSV format, with an optional Parquet version. Selected scripts and figures are available in this repository.

License This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments This work was supported by the Faculty of Engineering of the University of Santiago of Chile, Chile.

Contact For questions, please contact Claudio Urrea at claudio.urrea@usach.cl.