Autonomous Mobile Robot via Machine Learning - CS7641 - Machine Learning

The objective of this project is to develop autonomous, mobile robot ML algorithms that must navigate a Turtlebot3 through an environment to capture a care package

Table of Contents

• Features
• File Structure
• Changelog

Features

• Filtering and cleaning sensor data
• Object Detection
• Mapping
• Control

File Structure

• data: folder containing the data from the simulations
  • raw_dataset_lidar.csv: lidar sensor data
  • raw_dataset_odom.csv: encoder data
  • merged_lidar_odom_data.csv: merx`ged data
• features
  • radar_plot_frames: contains frames of the lidar data
  • plotting.ipynb: (Depricated) preprocessing data
• figs: Folder containing figures for the report
  • gantt.png: GANTT chart
  • objective_fig.png: Project objective graph
• preprocessing: Preprocessing of data
  • lidar_preprocessing.ipynb: Notebook containing the pipeline to generate the radar graph
  • cluster_objects.ipynb: Notebook that uses the raw lidar data and find the optimal parameters to cluster
  • icp.ipynb: Notebook that implements ICP algorithm
  • DBSCAN.py: Python script containing the DBSCAN class and functions
  • radar_plot.gif: Image of the simulation data from the LIDAR sensor
  • sample_DBSCAN_clustering.png: Figure of the sample DBSCAN clusters
  • optimal_DBSCAN_clustering.png: Figure of the optimal DBSCAN clusters
• reports:
  • figs: Folder containing figures for the report
    • gantt.png: GANTT chart
    • objective_fig.png: Project objective graph
  • main.pdf: PDF version of the report
  • main.tex: Latex file of the report
• src: folder containing the ML pipelines
  • ml_robotics_interfaces: ROS interfaces code
  • ml_robotics_project: ROS program files
    • ml_robotics_project: ROS program code
      • algorithm_handlers: Translate beteween ROS and ML
        • IPpoHandler.py: Interface for PPO handlers
        • ISlamHandler.py: Interface for SLAM handlers
        • IYoloHandler.py: Interface for YOLO handlers
        • PpoHandler.py: Handler for PPO algorithm
        • SlamHandler.py: Handler for SLAM algorithm
        • YoloHandler.py: Handler for YOLO algorithm
      • ml_algorithms: ML algorithms
        • Slam.py: AStar algorithm class
        • Yolo.py: Node that performs SLAM using lidar and odometry data
        • PPO2: All PPO nodes, algorithms, and data
      • ros_nodes: Classes that connect to the ROS environment
        • CameraNode.py: Node that captures camera images and feeds it into ROS
        • YoloNode.py: Node that computes objective coordinates given an image
        • INode.py: Initializes variables and parameters
        • PpoNode.py: Node that computes desired robot velocity given odometry and trajectory data
      • util: Folder containing ROS Utility scripts
        • ROS_message_utils: Constructs a ROS2 PoseStamped message
      • ppo_node_entry_point.py: This module implements the entry point for the PPO node
      • slam_node_entry_point.py: This module implements the entry point for the SLAM node
      • yolo_node_entry_point.py: Main entry point for the YOLO node
      • rel_pos_finder_node_entry_point.py: Entry point for the PosFinderNode
      • camera_node_entry_point.py: Entry point for the camera node.
  • object_detection: Object detection pipeline
    • predict: Script used for predition
    • yolo_training.py: Script used for Object detection training using YOLO
  • turtlebot3_example_code: Pre-build Turtlebot ROS program code for reference
  • q_learning.ipynb: Q-Learning for path planning and control
• rosbag_to_csv.py: Python script to generate the data from the simulation
• simulation.md: Readme to run a simulation

Changelog

All the changes in the repository will be documented here.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Version 0.3

This is the final deliverable version

Version 0.2

This is the midterm deliverable version

• Added data
• Preprocessing and Cleaned Data
• Deployed ML algorithms
• Generated initial results and graphs

Version 0.1

This is the initial push of our project.

• Completed the project proposal
• Initialized the project page on Github
• Compiled a guide to run simulations