GIMLI Installation and User Guide

GIMLI (Guided Integrated Maritime Logistics and Intelligence) is a software solution developed to navigation support for a semi-autonomous vessel. This guide outlines the necessary components, installation steps, configuration details, and instructions for running the solution in a simulated environment.

---

1. Introduction

1.1 Overview

• Purpose:
  GIMLI integrates LIDAR and camera sensor data to deliver real-time obstacle detection and classification for safe maritime navigation.

1.2 Key Objectives

• Implement reliable obstacle detection via sensor fusion.
• Synchronise real-time image and LIDAR data.
• Integrate and validate the system using the AILiveSim simulator.

---

2. System Components

2.1 Simulator: AILiveSim

• Description:
  AILiveSim provides a realistic maritime simulation environment built on Unreal Engine 4.
• License Requirement:
  A valid product key is required. Contact AILiveSim for licensing details.
• Documentation:
  Follow the instructions in the AILiveSim documentation.

2.2 Software Libraries

• Python Dependencies:
  All required Python libraries are listed in the requirements.txt file.
• Key Modules:
  • ALSLib: For sensor data acquisition and control.
  • Open3D & OpenCV: For 3D visualisation and image processing.
  • PyTorch & YOLOv5: For object detection within the camera feed.
  • Scikit-learn: For DBSCAN clustering.

2.3 Configuration Files

• Simulation Configuration:
  Underlying config files (e.g., sensor settings, extrinsic/intrinsic parameters) must be set up as required.
• User Modifications:
  Adjust file paths and sensor configurations within these files to match your environment.

2.4 Final Code

• Project Repository:
  Contains the complete implementation including:
  • Data acquisition threads.
  • Sensor fusion and calibration modules.
  • YOLO inference integration.
  • Point cloud clustering.
  • Object tracking.
  • Visualisation components.
• Code Location:
  The code is provided within this repository along with this guide.

---

3. Setup and Installation

3.1 Prerequisites

• Hardware:
  A system capable of running the simulator (ensure your system meets the hardware specifications as per AILiveSim) and process real-time sensor data.
• Operating System:
  Any OS that supports Python and the AILiveSim simulator.
• Software Requirements:
  • Python 3.11.9 or newer.

3.2 Installing AILiveSim

1. Obtain License:
   Contact AILiveSim to receive a valid product key.
2. Download & Install:
   Follow the AILiveSim documentation to install the simulator on your machine.
3. Activate Simulator:
   Place the license key in the approapriate folder and start the AILiveSim executable.

3.3 Setting Up the Python Environment

1. Clone the Repository:

   git clone https://github.com/belindbl/gimli.git
   cd gimli

2. Create a virtual environment:

   python -m venv venv

   Then activate the virtual environment:

   venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt

   This ensures all necessary dependencies are installed.

3.4 Configuring Simulation Files

• Edit Config Files:
  Locate and adjust the simulation configuration files (e.g., sensor settings, extrinsic/intrinsic parameters) as necessary.
• Verify Settings:
  Ensure that the sensor configuration matches the setup described in the project.

---

4. Executing the Final Code

4.1 Starting the Simulator

• Launch the AILiveSim executable.
• Ensure the simulator is running and that the desired scenario (e.g., 'ABoat') is loaded.

4.2 Launching GIMLI

1. Run the Main Script:
   Execute the final code to start the sensor fusion and visualisation processes (automatically loads a predetermined scenario):

python \GIMLI\Python\CustomScripts\GIMLI\RTDP9.py # RTDP9 is the current main script

2. Monitoring Output:

• A window should display the camera feed overlaid with the projected LiDAR points and annotation texts.
• A separate window will visualise the point cloud in real time.

3. User Interaction:
   Use the interface to monitor real-time data fusion and observe obstacle detection and collision avoidance overlays.