SORA-ATMAS: Adaptive trust management and multi-LLM aligned governance for future smart cities

Antuley, Usama; Siddiqui, Shahbaz; Hameed, Sufian; Arif, Waqas; Shah, Syed Attique

SORA-ATMAS: Adaptive trust management and multi-LLM aligned governance for future smart cities

Antuley, Usama and Siddiqui, Shahbaz and Hameed, Sufian and Arif, Waqas and Shah, Syed Attique (2026) SORA-ATMAS: Adaptive trust management and multi-LLM aligned governance for future smart cities. Knowledge-Based Systems, 337. p. 115403. ISSN 0950-7051

[thumbnail of 1-s2.0-S0950705126001462-main.pdf]

Preview

Text
1-s2.0-S0950705126001462-main.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.
Download (7MB)

Official URL: https://doi.org/10.1016/j.knosys.2026.115403

Abstract

The rapid evolution of smart cities relies on agentic AI for autonomous decision-making, yet introduces governance, risk, and compliance (GRC) challenges in decentralized environments. We propose SORA-ATMAS, an adaptive trust management and multi-LLM governance framework for smart-city disaster management. Evaluation with Weather, Traffic, and Safety agents shows the framework steers multiple LLMs (GPT, Grok, DeepSeek) toward policy-aligned outputs, reducing mean absolute error by 35% on average. Results demonstrate stable weather monitoring, effective handling of high-risk traffic plateaus (R ≈ 0.85), and adaptive trust regulation in safety scenarios. Runtime profiling confirms scalability, with throughput of 13.8–17.2 req/s, execution times  < 72 ms, and governance delays  < 100 ms for a 3-agent deployment; analytical projections indicate maintained performance at larger scales. Cross-domain policies ensure safe interoperability, such as allowing traffic rerouting only under validated weather conditions. SORA-ATMAS thus provides a regulation-aligned, verifiable governance framework that transforms distributed agent outputs into accountable, real-time decisions, offering a resilient foundation for smart-city management.

Item Type:	Article
Identification Number:	10.1016/j.knosys.2026.115403
Dates:	Date Event 22 January 2026 Accepted 2 February 2026 Published Online
Uncontrolled Keywords:	Adaptive trust management, Collaborative services, Agentic AI, GRC, Multi-LLM evaluation, Smart contracts, Smart-city governance, SDIoT
Subjects:	CAH11 - computing > CAH11-01 - computing > CAH11-01-01 - computer science
Divisions:	Architecture, Built Environment, Computing and Engineering > Computer Science
Depositing User:	Gemma Tonks
Date Deposited:	12 May 2026 13:11
Last Modified:	12 May 2026 13:11
URI:	https://www.open-access.bcu.ac.uk/id/eprint/17039