Event-based LEarning, Vision, And conTrol for Embodied systems

• Unité de recherche: SEFRI
• Numéro de projet: 25.00438

Résumé des résultats (Abstract) (Anglais)

Computing with neuromorphic hardware, learning with spiking neural networks, and sensing with event-based cameras are recent examples of new technologies that have the potential of revolutionizing the learning and control capabilities of digital automation and robotics. They offer a novel and unique opportunity to make machine intelligence more efficient, more embodied, and more resilient. Event-based technology will enable online learning over multiple temporal and spatial scales, sensors with embodied adaptive intelligence, and improved translation of rhythmic features of animal intelligence in machines. Yet the discipline is still in its infancy and technology lies ahead of methodology. While the potential of event-based information processing has been researched separately in neuroscience, in algorithmic science, and in electronic circuit design, the deployment of event-based machine intelligence across the broad spectrum of automation technologies requires a much higher level of integration. The sparse and asynchronous nature of event- based information impacts the design all the way from the sensing and computing hardware to the perception and decision-making functionalities of the entire system. The Doctoral network ELEVATE proposes to educate a new generation of information engineers able to address those challenges through a unique environment combining leading international recognition in complementary scientific disciplines, a diverse range of companies at the forefront of event-based technology, and a culture of machine design inspired by the challenge of further digitalizing our societies with a focus on sustainability, security, and ethics.