Description du projet
Des indices radio environnementaux pour soutenir la prise de décision des agents autonome
Les signaux radio, qui utilisent des capteurs de radiofréquences, apportent des informations électromagnétiques avec une signification contextuelle. Ils permettent aux agents de prédire les résultats et de naviguer plus efficacement dans les environnements sociaux. L’interprétation des signaux radio contextuels implique un échange d’informations qui tienne compte de la capacité des autres agents à traiter les stimuli environnementaux. Le projet CUE-GO, financé par le CER, est à l’origine d’un nouveau cadre qui permettra aux agents autonomes d’améliorer la prise de décision grâce à l’utilisation d’indices radio environnementaux. Cela comprend le développement de méthodes d’extrapolation des signaux radio, d’algorithmes de navigation guidée par signaux et d’évaluations pour les prédictions futures. S’inspirant de divers domaines, dont les neurosciences comportementales et l’ingénierie, le projet fera progresser la prise de décision inspirée par l’homme pour les agents autonomes. CUE-GO a pour ambition de contribuer à une société dans laquelle les humains et l’IA coexisteraient de manière transparente.
Objectif
The CUE-GO project aims to conceive a novel methodological framework for enhancing the decision-making capabilities of autonomous agents through the exploitation of contextual radio cues of the environment. Radio cues represent a quantum leap from the traditional concept of features, usually retrieved by vision-based systems, as they contain electromagnetic information with semantic meanings (contextual) enabled by radio-frequency sensors, like those at TeraHertz bands. The elaboration of contextual radio cues allows a far-reaching prediction of the outcomes of agents’ behaviors and ultimately yields a more efficient navigation in social environments, more accurate localization of people and objects, and enhanced cooperation toward a common mission goal. In interpreting contextual radio cues, agents exchange their sensed information in a way that considers other agents’ expertise, i.e. their abilities to process environmental stimuli.
To achieve this vision, I will: (1) develop a general framework for the decision-making of autonomous agents that emulates the human capability of interpreting cues for anticipating an action’s course; (2) conceive and design methods for extrapolating contextual radio cues based on high-resolution semantic mapping of the environment; (3) conceive and design cue-guided localization and navigation algorithms that will boost ambient awareness; (4) conceive new methods and metrics to assess the agents’ skills in associating contextual radio cues with statistical models that accurately predict future rewards or punishments; (5) develop collaboration schemes accounting for the assessment of agents’ expertise.
Thanks to a multidisciplinary approach, combining diverse knowledge from behavioral neuroscience to engineering, this project will lead to a significant advance in human-inspired decision-making for future networks of autonomous agents, toward a society where humans and artificial intelligence co-exist in the same environment.
Champ scientifique
- natural sciencesbiological sciencesneurobiology
- natural sciencescomputer and information sciencesartificial intelligence
- engineering and technologyelectrical engineering, electronic engineering, information engineeringinformation engineeringtelecommunicationsradio technologyradio frequency
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringsensors
- natural sciencesmathematicsapplied mathematicsstatistics and probability
Mots‑clés
Programme(s)
- HORIZON.1.1 - European Research Council (ERC) Main Programme
Thème(s)
Régime de financement
HORIZON-ERC - HORIZON ERC GrantsInstitution d’accueil
00185 Roma
Italie