CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Diseño e implementación de un UGV para la deteccion de radiación gamma

Estudiants que han llegit aquest projecte:

Director/a: ROYO CHIC, PABLO

Departament: DAC

Títol: Diseño e implementación de un UGV para la deteccion de radiación gamma

Data inici oferta: 02-02-2025     Data finalització oferta: 02-10-2025


Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: EETAC
 
Segon director/a (UPC): SÁNCHEZ SEGURA, ALBERT
 
Paraules clau:
UGV, Radiación Gamma, Detection, CNN, AI, IA, inteligencia artificial, segmentación de imágenes, detección de radiación gamma, GammaDrone, navegación autónoma, visión por computador, U-Net, PyTorch, ArduPilot, ArduPilog
 
Descripció del contingut i pla d'activitats:
El estudiante tendrá que realizar el diseño de un rover autónomo equipado con un detector gamma para realizar un mapa de contaminación de un área determinada.

Las tareas a realizar son:

1.- Diseño de un UGV equipado con un sistema de ardurover.
2.- Realizar la calibración del rover para mejorar la navegación del rover y demostrar que el rover es capaz de seguir un plan de vuelo de forma correcta.
3.- Instalar el sistema gammaDrone que utilizamos en los drones y un detector gamma para realizar medidas a nivel de suelo.
4.- Documentar el trabajo realizado
 
Overview (resum en anglès):
This Project presents the development and implementation o fan unmanned ground vehicle (UGV) equipped with autonomous navigation, computer vision based on artificial intelligence, and gamma radiation detection, within the framework of the GammaDrone Project by the Spanish Nuclear Safety Council (CSN). It acts as a complementary alternative to unmanned aerial systems (UAS), specially in those environments where this systems could encounter operational limitations.A cost-effective, robust, and modular rover has been built to assess its potential in emergency response or inspection contexts.
The system integrates both hardware (chassis, sensors, video and telemtry transmitters, receiver) and software components (ArduPilot firmware, MAVLink protocol, Mission Planner interface), enabling autonomous or manual missions via radio control or intuitive peripherals like a steering wheel and pedals. It incorporates advanced elements such as LiDAR, GNSS positioning, digital video transmisión, and radiation sensing using the SiPM-1000 detector. A convolutional neural network was trained and deployed onboard to perform real-time semantic segmentation of the terrain and adapt the rover's cruise speed according to the type of terrain in front of the vehicle (although the system has more untested potential).
The project follows an iterative and multidisciplinary approach, covering mechanical design, electronic integration, software configuration, and artificial intelligence understanding, training and testing. Overall vehicle testing was conducted in various configurations, speeds and terran types to validate system stability and the model's segmentation performance. The neural network achieved high accuracy, over 90% in well-represented environments, though performance dropped in reflective surfaces like wáter due to dataset limitations.
The conclusions validate the feasibility of using the rover as a functional platform for radiation data collection tasks. Furthemore, posible improvements are identified base don the results obtained from various tests, evaluating the system's dynamic behavior in real-world scenarios. Additionally, the development and progression of the project itself are analyzed.

© CBLTIC Campus del Baix Llobregat - UPC