Year: 2025
Type: peer-review
Abstract: High Level: This work extends the concept of Vertical AI Agents to "Physical AI Agents", embodied AI with LLM cognition.The author proposes a standardized modular architecture for these Physical AI Agents with three core components (perception, cognition, and actuation) and introduces Physical Retrieval Augmented Generation (Ph-RAG) to connect physical intelligence with industry-speci c LLMs.Case studies demonstrate applications across multiple sectors, including oil and gas pipeline monitoring and warehouse automation.Detailed: The author rst motivates the need for vertical AI agents.Since AGI is still undeveloped, the application of LLMs to speci c complicated application verticals requires model ne-tuning and modi cation.They claim that adding in application-speci c context, thereby creating "Vertical AI Agents", can enhance performance in elds such as supply chain management, the oil and gas
Source: Qeios
DOI: 10.32388/ugymvs
Author: Bahram Jalali
Institution: University of California, Los Angeles
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4408057236