This repository contains the Hexapod Caterpie project, developed as part of the subject Artificial Intelligence Applied to Robots in the Computer Engineering Degree at the University of Huelva. The main objective of the project was to design, implement, and configure a hexapod robot capable of detecting and recognizing nanobugs in real-time.
- Introduction
- Project Features
- Requirements
- Installation
- Usage
- Project Architecture
- Challenges and Solutions
- Future Work
- Authors
- License
This project involves the implementation of a hexapod robot based on the FreeNove Big Hexapod model. The robot uses a YOLOv5 neural network to detect specific objects such as nanobugs of different colors and other hexapods. Additionally, it integrates a communication system between robots to coordinate actions.
The primary goal was to develop the following capabilities:
- Autonomous movement avoiding surface edges.
- Real-time detection and capture of nanobugs.
- Communication between robots to coordinate and avoid conflicts in capturing nanobugs.
- Autonomous Movement: Implementation of algorithms for the robot to move while avoiding edges using distance sensors.
- Nanobug Detection: Utilization of the YOLOv5 model to identify nanobugs and other hexapods.
- Inter-Robot Communication: Centralized system using a server to coordinate nanobug capture and avoid interference.
- Decision Algorithms: Logic based on edge detection, nanobug detection, and hexapod detection to make real-time decisions.
-
Hardware:
- FreeNove Big Hexapod Robot Kit.
- Raspberry Pi (recommended model: 3B+ or higher).
- Compatible servomotors.
- Distance sensor (sonar).
-
Software:
- Python 3.8 or higher.
- Libraries:
- PyTorch
- OpenCV
- Pyro4
- RoboFlow (for dataset management)
- YOLOv5 (official Ultralytics implementation).
-
Clone this repository:
git clone https://github.com/GrunCrow/Hexapod_Caterpie.git cd Hexapod_Caterpie -
Install dependencies:
pip install -r requirements.txt
-
Configure the environment on the Raspberry Pi and ensure the servomotors are correctly connected.
-
Download the trained YOLOv5 model into the
models/folder.
-
Run the main script to start the robot:
python main.py
-
Set up communication between robots if necessary:
python server.py
- Movement: The robot moves forward, detects edges, and turns accordingly to avoid falls.
- Object Detection: Captures real-time images and uses YOLOv5 to identify nanobugs.
- Coordination: Implements a centralized protocol to avoid conflicts in nanobug capture.
- Movement Control: Base code for calibration, turns, and robot displacement.
- Edge Detection:
- Use of sonar to measure distances and avoid falls.
- Logic to recalibrate movement after detection.
- Nanobug Recognition:
- YOLOv5 model trained with labeled images of nanobugs and hexapods.
- Real-time image processing from the robot's camera.
- Inter-Robot Communication:
- Client-server system based on Pyro4.
- Robot Assembly: Error in the placement of the battery compartment, resolved by adjusting the assembly.
- Servo Overheating: Limiting the head's rotation angle to avoid collisions and damage.
- Damaged Leg Servo: Replacement of the defective component.
- Actions on Detecting Other Hexapods: Implement additional behaviors such as reversing and turning.
- Improved Training: Increase dataset quality to avoid confusion between hexapods and the background.
- Edge Detection Optimization: Use depth estimation models to improve accuracy.
- Additional Sensors: Incorporate more precise sensors like laser sensors.
- Juan Diego Díaz
- Alberto Fernández Merchán
- Alba Márquez Rodríguez
This project is open source under the MIT License, unless otherwise stated. For more details, see the LICENSE file.