Campus

Artificial vision and deep learning

Deep Learning for industrial image processing, obstacle detection and navigation

Development of computer vision algorithms for use in smart mobility, predictive maintenance and industrial mobile robotics.

  • 2023 - 2025
  • Finished
  • Basque Country
  • Funding: Mobility Lab Foundation

What is the project?

The project develops deep learning algorithms for industrial image processing, obstacle detection and vision-based intelligent navigation. The work is structured around three main areas:

- Development of navigation algorithms based on vision and deep neural networks.

- Recognition of natural landmarks based on elements in the environment, without the need to modify it with specific markers.

- Optimisation of trajectories, energy consumption and predictive maintenance by calculating the remaining service life of vehicles.

The challenge: computer vision for smart mobility

Autonomous and industrial mobility systems need to interpret their surroundings, navigate accurately and anticipate faults without relying on artificial infrastructure.

  • Navigation without disrupting the environment

    The challenge is to locate and guide vehicles using natural features in the environment, without installing specific markers.

  • Obstacle and distance detection

    Indoor autonomous vehicles need to detect objects and estimate distances reliably.

  • Predicting breakdowns

    Predictive maintenance involves estimating the remaining service life in order to determine when a vehicle may become unavailable.

  • Trajectory and energy optimisation

    The routes taken by robots and industrial vehicles must reduce energy consumption and adapt to real-world operating conditions.

Project approach

The project combines computer vision, deep neural networks and industrial collaboration to develop intelligent systems for navigation, detection and predictive maintenance.

  • Deep Learning applied to industrial images
  • Navigation based on vision and natural patterns
  • Optimisation of RUL, flight paths and energy consumption

This approach enables the integration of doctoral research, scientific validation and technology transfer to local businesses:

  • We identify natural fiducial patterns without altering the physical environment
  • We develop navigation and detection algorithms for autonomous vehicles
  • We estimate the remaining service life of vehicles to improve maintenance
  • We are seeking new partnerships in industrial AI and mobile roboticsl

Project objectives

To develop deep learning algorithms for computer vision applied to intelligent navigation, obstacle detection and predictive vehicle maintenance.

  • 1

    Develop vision-based navigation

  • 2

    Identifying natural fiducial patterns

  • 3

    Processing industrial images using deep learning

  • 4

    Optimising maintenance using RUL

  • 5

    Improving performance and energy consumption

Scientific output and knowledge transfer

Applied research in computer vision, autonomous navigation and predictive maintenance

Indexed publications

Deep learning, navigation and autonomous vehicles

The project has resulted in three publications in indexed journals on robotic localisation using natural patterns, navigation using reinforcement learning and genetic algorithms, and object detection with distance estimation for indoor autonomous vehicles.

Transfer in progress:

  • Stirling Centre / MC3

    Optimisation of vehicle RUL and predictive maintenance using deep neural networks.

  • Aldakin

    A collaboration aimed at optimising energy consumption in the routes of industrial mobile robots.

The solution: AI for vision and intelligent navigation

A set of deep learning algorithms for interpreting images, detecting obstacles, navigating and anticipating faults in industrial mobility systems.

Areas of development:

  • Vision-based navigation

  • Natural fiduciary models

  • Object and distance detection

  • Optimisation of the RUL

  • Efficient routes

Technological capabilities:

  • Tracks robots using environmental patterns without artificial markers

  • It recognises natural visual cues to aid navigation.

  • It detects objects and estimates distances in indoor autonomous vehicles.

  • Calculate the remaining service life to anticipate breakdowns.

  • It promotes more efficient routes and lower energy consumption.

Team and collaborations

A project focusing on industrial AI, computer vision, autonomous navigation and predictive maintenance, with ongoing doctoral research and industry partnerships.

Funding

Project funded by Mobility Lab Vitoria-Gasteiz
Implementation period: 2023–2025