AI-Based Assistance System for Control Rooms in Large-Scale Infrastructures

Agentic AI for Safety-critical Multi-drone Systems: Challenges and Opportunities

Multi-drone systems are increasingly positioned for safety-critical missions such as search and rescue (SAR) and critical infrastructure monitoring. Yet, real-world adoption remains constrained not only by autonomy performance, but by the difficulty of integrating agentic behavior into professional work: operators must understand, trust, and govern automation under uncertainty, time pressure, and accountability. This position paper synthesizes the ambitions and lessons from two ongoing efforts: NAMUR, which explores LLM-supported robot control in SAR and firefighting contexts, and PERSIST, which explores persistent drone operations for monitoring and security at critical infrastructure sites. We argue that agentic AI should be approached as a socio-technical design problem, where interfaces, oversight mechanisms, and evaluation practices are as critical as algorithms. We outline a human-centered, participatory, and iterative research approach aimed at uncovering stakeholder needs, shaping agent capabilities through successive prototypes, and producing transferable proof-of-concept systems and evaluation strategies for other safety-critical contexts.

Modular OODA-Based Agents for Scalable and Adaptive AI in System-of-Systems Modelling and Simulation

Abstract. Agent-Based Modelling and Simulation (ABMS) is widely used in System-of-Systems (SoS) studies to represent constituent systems operating in dynamic environments. However, the absence of standardised agent architectures hinders scalability, behaviour traceability, and comparative evaluation of emergence, coordination, and operational performance. This work introduces a modular agent design based on Observe–Orient–Decide–Act (OODA) loops as a framework for SoS simulations, enabling transparent information flow and Artificial Intelligence (AI) integration at the decision layer. A wildfire suppression scenario is used for the study, modelling firefighting crews, aircraft, helicopters, bulldozers, and an incident commander as OODA-driven agents. Results show that OODA-based agents exhibit coherent and traceable team behaviours, adaptive task switching, and communication-driven coordination, supporting the study of emergence and interoperability in SoS operations. The Decision stage is designed for future improvements in the form of learning-based AI and Large Language Models to enhance autonomy and operational fidelity.

Network Orchestration Framework Design Using AI-Driven Automation and Cybersecurity

This paper addresses human error in network orchestration systems and the high cost and resource requirements of integrating artificial intelligence (AI) for network orchestration. It proposes a framework for implementing an AI decision-maker and automation. Data are fed into the AI decision-maker to trigger designated automation robots’ tasks or notify IT specialists to gradually implement automated robots, ensuring efficient resource use, reducing costs, and enhancing productivity. We evaluated the proposed method in a simulation with genuinely uncertain outcomes, across 20 independent runs: the AI decision-maker reached 78.3% accuracy against an estimated 79.1% achievable ceiling, and the proposed framework reduced operational cost by 61.4 ± 0.7% relative to fully manual operation—the best of six operating policies in the training environment—while an explicit sensitivity guard, rather than the learned model, accounts for the absence of security incidents; under distribution shift, the framework retains 43.2 ± 0.8% savings, second only to a hand-tuned rule-based router that requires environment-specific threshold calibration. However, the proposed method requires an IT specialist to implement it properly, and the AI model’s accuracy depends on the amount of input data. In the end, we recommend that future work conduct a study focused on AI decision-makers, test the proposed method on real-world companies, and implement AI decision-makers across various departments to cover a broader range of the company’s systems.

LLM-Based Agents for Software and Systems Security: Approaches, Applications, and Assessment

A systematic literature review of technical approaches, including agent architecture, perception, memory, reasoning and planning, action space, orchestration, and self-improvement, reveals a field that has built agents able to act but not yet agents whose authority is bounded or whose behavior is auditable.

LLMs and Agentic AI Systems for Smart Grids: A Tutorial on Architectures and Applications

A solver-grounded design principle is presented: a numerical result is reported only when it originates from a trusted tool and passes explicit verification, and a four-group evaluation framework spanning task utility, solver-grounded correctness, faithfulness and safe failure, and cost and latency is proposed.

Agent-in-the-Loop: Using AI Agents to Perform Control-Oriented Simulation Tasks