Real-Time Obstacle Avoidance and Navigation Using Depth Estimation for Autonomous UAVs

A Python-based implementation for enhanced Tello drone control, integrating UWB (Ultra-Wideband) positioning and depth estimation. This project is designed to enable precise navigation, real-time obstacle detection, and visualisation of the drone's environment.

---

Features

• UWB Position Tracking:

  • Real-time drone position tracking using the Nooploop Linktrack system.
  • ROS-based streaming and integration for seamless communication.

• Depth Estimation:

  • Utilises Zoe depth mapping to generate accurate depth information from stereo images.
  • Supports clustering and segmentation for obstacle detection.

• Obstacle Avoidance:

  • Processes depth maps to dynamically map obstacles and ensure safe navigation.

• Real-Time Visualisation:

  • Displays drone position and detected obstacles in a graphical interface.

• Task Logging:

  • Logs drone configurations and position data for debugging and analysis.

---

System Requirements

• Hardware:

  • DJI Tello drone.
  • Nooploop Linktrack UWB system.

• Software:

  • Ubuntu 20.04
  • Python 3.8
  • ROS1-Noetic
  • Required Python libraries:
    • numpy
    • opencv-python
    • matplotlib
    • torch
    • PyAV
    • Additional dependencies listed in requirements.txt.

---

Installation

1. Clone the repository:

   git clone https://github.com/horse-3903/NTU-UAV-Research.git
   cd tellodrone-project

2. Install Python dependencies:

   pip install -r requirements.txt

3. Configure ROS:

   • Install and set up ROS on your system.
   • Ensure compatibility with Nooploop Linktrack and Tello drone SDKs.

4. Add calibration data:

   • Place calibration_data.nps in the project root, containing the camera matrix and distortion coefficients.

---

Usage

1. Start the UWB System

Run the UWB initialisation script to begin position tracking:

bash cmd/uwb.sh

2. Verify UWB Data

Ensure UWB data is streaming correctly via ROS:

rostopic echo <nlink_linktrack_nodeframe1>

3. Launch the Main Task

Run the core script to execute tasks:

python task/main.py

---

Key Features in Detail

Position Tracking

• Accuracy: Tracks the drone's position in real-time using UWB and ROS.
• Integration: Dynamically updates positions for effective visualisation and control.

Depth Mapping

• Generates high-accuracy depth maps using Zoe depth estimation.
• Segments depth clusters to identify obstacles in the environment.

Visualisation

• Displays annotated video streams with obstacle information and dimensions.
• Offers real-time updates on drone actions and environment mapping.

Logging

• Position Logs: Records the drone’s position at each step.
• Configuration Logs: Captures:
  • Takeoff and target positions.
  • Detected obstacles.
  • Current drone configurations.

---

Future Enhancements

1. Dynamic Re-Routing:

   • Advanced algorithms for re-routing in complex environments.

2. SLAM Integration:

   • Combining UWB and visual SLAM for improved localisation.

3. Machine Learning:

   • Predictive obstacle avoidance using ML models.

---

Contributors

• Your Name: GitHub Profile

Feel free to contribute by submitting pull requests or opening issues for bugs or feature suggestions.

---

License

This project is licensed under the MIT License.

---