Product Design, Manufacturing & Innovation Resources

Dernières publications et brevets sur l'IA agentique

Agent IA

Astuce : suite à la sélection ci-dessous, vous pouvez rechercher et filtrer nos 2 bases de données complètes :

> outil de recherche de publications gratuit < par auteur, sujet, mots-clés, date ou revue.

> outil gratuit de recherche de brevets < pour les brevets en anglais de l'Office européen des brevets.

Vous êtes un agent.
L'IA agentique révolutionne la conception de produits en permettant une prise de décision autonome et orientée vers un objectif grâce à des processus itératifs.

L'IA agentique désigne les systèmes d'IA qui poursuivent des objectifs en plusieurs étapes de manière autonome en itérant à travers des boucles de perception, de planification, d'utilisation d'outils et d'auto-correction sans nécessiter d'intervention humaine à chaque point de décision — une rupture structurelle avec les modèles de réponse instantanée à un seul tour vers des architectures qui décomposent les objectifs complexes en sous-tâches, exécutent des actions sur des environnements externes, évaluent les résultats et révisent les plans en conséquence.

L'architecture agentique canonique associe un grand modèle linguistique de raisonnement à un registre d'outils - navigateurs web, interprètes de code, clients API, systèmes de fichiers, interfaces de base de données - et à une architecture de mémoire couvrant le contexte de travail à court terme, les enregistrements épisodiques des actions passées et les connaissances à long terme récupérées, ce qui permet à l'agent de poursuivre des objectifs cohérents sur des horizons d'interaction étendus qui dépassent toute fenêtre de contexte unique. Les configurations multi-agents vont plus loin, distribuant les sous-tâches à des agents spécialisés coordonnés par un orchestrateur, introduisant des fonctions inter-agents, des fonctions de gestion de l'information, des fonctions de gestion de l'information et des fonctions de gestion de l'information. communication les protocoles, l'agrégation des résultats et la résolution des conflits sont des préoccupations architecturales supplémentaires.

The publications and patents indexed below address planning algorithms, tool use architectures, memory systems, multi-agent coordination protocols, agent evaluation benchmarks, and safety constraint méthodologies.

Voici notre dernière sélection de publications et de brevets mondiaux en anglais sur l'IA agentique, parmi de nombreuses revues scientifiques en ligne, classées et axées sur l'IA agentique, l'agent d'IA, l'agent d'IA autonome, le système multi-agent, l'orchestration de l'agent d'IA, la boucle de planification de l'agent, l'agent ReAct, la planification de la chaîne de pensée, l'agent d'IA utilisant des outils, l'appel de fonction à l'IA, la mémoire de l'agent d'IA, la mémoire à long terme de l'agent, la mémoire épisodique de l'agent, mémoire de travail de l'agent, décomposition des tâches de l'agent d'IA, planification hiérarchique de l'agent, autocorrection de l'agent d'IA, réflexion de l'agent d'IA, évaluation de l'agent d'IA, mise en forme de la récompense de l'agent, bac à sable de l'agent d'IA, garde-fous de sécurité de l'agent, coordination multi-agent, protocole de communication de l'agent, registre des outils de l'agent d'IA, agent exécutant le code, agent navigant sur le web, agent augmenté par la recherche, évaluation de référence de l'agent et agent humain dans la boucle.

Ai agent for core network

Patent published on the 2026-10-15 in WO under Ref WO2026213740 by NOKIA TECH OY [FI] (Gajic Borislava [de], Sun Shu Qiang [cn], Abdelkader Abdelrahman [de], Khatibi Sina [de], Schirilla Achill Andreas [de])

Abstract: Example embodiments of the present disclosure are directed to artificial intelligence (AI) agent for core network. A first network entity receives, from a second network entity or an apparatus, a request originated from the apparatus for an agentic service. The first, second network entities and the apparatus implement an AI agent. The first network entity determines a plan for the agentic service at least based on capability information of at least one device. The plan is associated with one or[...]


Our summary: An AI agent for core networks processes requests for agentic services. It determines a service plan based on device capabilities and executes associated tasks. The execution status is communicated back to the requesting entity.

AI agent, core network, agentic service, execution status

Patent

Ai agent and generative ai-based b2b solution comparison, recommendation, and deployment automation system and method

Patent published on the 2026-10-08 in WO under Ref WO2026212264 by IMPACT FLOW INC [KR] (Ryu Hyokwon [kr])

Abstract: Disclosed is a B2B solution deployment automation apparatus and a method that use an AI agent and a generative AI model to generate user requirement-based B2B solution recommendations and comparison reports, and furthermore, automatically orchestrate a deployment process such as integration configuration, API connection, and data migration, and provide operational optimization feedback through calculation of performance, cost, and security indices.[...]


Our summary: The system automates B2B solution deployment using an AI agent and generative AI model. It generates user requirement-based recommendations and comparison reports. The method also orchestrates integration, API connections, and data migration while providing optimization feedback.

B2B solutions, AI agent, generative AI, deployment automation

Patent

Secured distributed network context management for ai agent

Patent published on the 2026-10-08 in WO under Ref WO2026209409 by NOKIA TECH OY [FI] (Aggarwal Chaitanya [de], Gajic Borislava [de], Tang Haitao [fi], Mishra Rajendrakumar Premnarayan [in])

Abstract: Example embodiments of the present disclosure are directed to secured distributed network context management for an artificial intelligence (AI) agent. A method comprises verifying, based on a context distribution policy associated with a second apparatus comprised in the at least one apparatus, that a network context is authorized to be provided to the second apparatus, wherein the context distribution policy indicates identifier information of the second apparatus and an indication of a networ[...]


Our summary: The method verifies network context authorization based on a context distribution policy. It identifies the second apparatus and the allowed network context. The authorized context is transmitted to the second apparatus for a generative AI agent.

secured distributed network, context management, artificial intelligence, policy verification

Patent

Collaborative and cooperative processing method for tasks in wireless network system, wireless network system, electronic device and computer program

Patent published on the 2026-10-08 in WO under Ref WO2026208347 by ZTE CORP [CN] (Yang Li [cn], Sun Wenwen [cn], Xie Feng [cn])

Abstract: Provided in the embodiments of the present disclosure are a collaborative and cooperative processing method for tasks in a wireless network system, a wireless network system, an electronic device and a computer program product, wherein the method is applied to a network element node, a communication-computing service layer and an AI agent (AA) task layer being deployed on the network element node, and the AA task layer comprising one or more AA functional entities. The method comprises: the AA t[...]


Our summary: The method enables collaborative processing for tasks in a wireless network system. It involves interaction between an AI agent task layer and a communication-computing service layer. This approach addresses efficient cooperation between different network element nodes across various vendors.

collaborative processing, wireless network, AI agents, task cooperation

Patent

System and method for accelerating operations of an ai agent on a digital interface

Patent published on the 2026-10-08 in US under Ref US20260311007 by EVINCED INC [US] (Synett Yossi [il], Moav Gal [il], Thadani Navin R [us], Or Evyatar Yaakov [il], Hakman Tair [il], Orgil Kfir Yehuda [il], Bracha Ben [il])

Abstract: [0000] The method and system for accelerating operations of an AI agent is presented. The method includes receiving, by a first AI agent, a task to be performed on at least one digital interface based on a natural language request from a client device; extracting, by the first AI agent, at least one learned skill to be used by the first AI agent for completing the task based on accessibility semantics of at least one component of the at least one digital interface, wherein learned skills are cap[...]


Our summary: The system accelerates AI agent operations on digital interfaces. It involves receiving tasks from client devices and extracting learned skills based on accessibility semantics. The AI agent completes tasks using these skills effectively.

AI agent, digital interface, natural language processing, learned skills

Patent

Novel edge-based multi-agent system for real-time data validation and quality grading

Patent published on the 2026-10-08 in US under Ref US20260312968 by PARA RAGHU KUL THILAK [CA] (Para Raghu Kul Thilak [ca])

Abstract: [0000] present invention relates to an edge-based multi-agent data validation system that executes specialized checks (schema, domain, semantic, anomaly) on incoming data streams in near real time. A coordinator module aggregates validation outputs and assigns a quality tier label (bronze, silver, or gold) to each data unit. This immediate classification prevents contaminated data from infiltrating downstream pipelines, thereby reducing bandwidth usage and storage overhead. Agents autonomously a[...]


Our summary: The invention presents an edge-based multi-agent system for real-time data validation. It performs specialized checks on incoming data streams and assigns quality tier labels. The system enhances data integrity and reduces resource usage in distributed environments.

multi-agent system, data validation, real-time processing, edge computing

Patent

A Scoping Review

Published on 2026-02-01 by Jonathan Gibson, Praveen Chinniah, Shashank Chapala, Ojasvi Vemuri, Rajesh Botchu @MDPI

Abstract: Objectives: Artificial intelligence (AI) is a transformative development in the field of medicine. In the field of musculoskeletal radiology, agentic AI is a technology that could flourish, but currently, the limited evidence base is fragmented and sparse, and we present a scoping review of it. Methods: Parallel searches were conducted in four databases: PubMed, Embase, Scopus, and Web of Science. Search terms included all agentic AI and autonomous AI agents, as well as radiology. All papers und[...]


Our summary: This scoping review evaluates the potential of agentic AI in musculoskeletal radiology. It identifies eleven relevant studies highlighting improved decision support, workflow optimization, and image analysis. Despite promising findings, the evidence base remains limited and theoretical.

AI in Radiology, Musculoskeletal Imaging, Workflow Optimization, Decision Support

Publication

A Novel Security Index for Assessing Information Systems in Industrial Organizations Using Web Technologies and Fuzzy Logic

Published on 2026-01-29 by Sulieman Khaddour, Fares Abu-Abed, Valery Bogatikov @MDPI

Abstract: Industrial information systems based on web technologies (ISOWT) face escalating security challenges, particularly in critical sectors such as energy. Traditional qualitative security assessments often lack the ability to deliver actionable, real-time insights for managing complex, dynamic threats. This paper proposes a novel security index for evaluating ISOWT in industrial organizations by integrating fuzzy logic, metric-based evaluation, fuzzy Markov chains, and multi-agent systems. The propo[...]


Our summary: This paper introduces a novel security index for evaluating industrial information systems using web technologies and fuzzy logic. It quantifies deviations from an ideal safety center, enabling early risk detection and proactive mitigation. The methodology shows significant improvements in security metrics and outperforms existing assessment techniques.

Security Index, Fuzzy Logic, Information Systems, Web Technologies

Publication

Sujets abordés : IA agentique, prise de décision autonome, processus itératifs, objectifs à plusieurs étapes, perception, planification, utilisation d'outils, boucles d'autocorrection, grand modèle de langage, architecture de mémoire, configurations multi-agents, protocoles de communication inter-agents, algorithmes de planification, critères d'évaluation des agents, méthodologies des contraintes de sécurité, résolution de conflits, orchestrateur ISO/IEC 27001, ISO/IEC 25010, ISO 9241, ISO/IEC 2382, et IEEE 1012.

Glossaire des termes utilisés

Application Programming Interface (API): un ensemble de règles et de protocoles qui permet à différentes applications logicielles de communiquer et d'interagir entre elles, permettant l'intégration de fonctionnalités et l'échange de données entre les systèmes.

Large Language Model (LLM): Modèle statistique entraîné sur de vastes ensembles de données textuelles pour comprendre et générer un langage proche de celui de l'humain. Il utilise des techniques d'apprentissage profond, notamment des réseaux de neurones, pour prédire les séquences de mots et effectuer diverses tâches de traitement automatique du langage naturel.

Contexte historique

2000
2002
2010
2013
2000
2000
2003
2010
2013-09-24

(si la date est inconnue ou non pertinente, par exemple « mécanique des fluides », une estimation arrondie de son émergence notable est fournie)

10e année anniversaire

Les images en pleine résolution et les téléchargements sont uniquement disponibles, et 100% gratuits, pour les membres inscrits.