Hybrides Ki-Orchestrierungssystem und Verfahren zur Verteilung von Aufgaben der Künstlichen Intelligenz über Vorrichtung, Netzwerk Und/oder Cloud-Ressourcen Hinweg
Anmelder: Deutsche Telekom AG 🇩🇪
Details
- Veröffentlichungs-Nr.
- EP4752721
- Anmeldetag
- 29. November 2024
- Veröffentlichung
- 3. Juni 2026
- Rechtsraum
- EP
- IPC
- G06F9/50, G06N20/00, H04L67/10
Abstract
A computer-implemented method for distributing an artificial intelligence, Al, task in a hybrid AI orchestration system, the method comprising: • initiating an AI request for an AI task at a UE-hybrid-AI-orchestrator associated with user equipment, UE; • transmitting a request from the UE-hybrid-AI-orchestrator to a network-assisted-Ai-orchestrator located in a telecommunications network, the AI request soliciting network-related information and/or remote AI capability information relevant to the AI task; • receiving, at the UE-hybrid-AI-orchestrator, network-related information and/or remote AI capability information from the network-assisted-Ai-orchestrator; • analyzing, by the UE-hybrid-AI-orchestrator, the received network-related information and/or remote AI capability information to determine whether to execute the AI task on the UE on an on-device AI model, on a remote AI model within the telecommunications network, and/or on a cloud-based AI model in a remote cloud environment; • executing the AI task based on the determination made by the UE-hybrid-AI-orchestrator, wherein the UE-hybrid-AI-orchestrator: ∘ transmits the AI task to an on-device AI model if local execution is determined, transmits the AI task to the distributed AI model if execution within the telecommunications network is determined, and/or transmits the AI task to a cloud-based AI model in a remote cloud environment if remote cloud execution is determined.
Anmelder
- Firma
- Deutsche Telekom AG
- Land
- 🇩🇪 Deutschland
Deutsches Telekommunikationsunternehmen mit Sitz in Bonn. Bietet Festnetz- und Mobilfunkdienste, Internetzugang sowie IT- und Netzwerklösungen für Privat- und Geschäftskunden im In- und Ausland an.
2.494 Patente in unserer Datenbank