Modelling and simulation of complex systems: an approach based on multi-level agents

A complex system is made up of many components with many interactions. So the design of systems such as simulation systems, cooperative systems or assistance systems includes a very accurate modelling of interactional and communicational levels. The agent-based approach provides an adapted abstraction level for this problem. After having studied the organizational context and communicative capacities of agentbased systems, to simulate the reorganization of a flexible manufacturing, to regulate an urban transport system, and to simulate an epidemic detection system, our thoughts on the interactional level were inspired by human-machine interface models, especially those in “cognitive engineering”. To provide a general framework for agent-based complex systems modelling, we then proposed a scale of four behaviours that agents may adopt in their complex systems (reactive, routine, cognitive, and collective). To complete the description of multi-level agent models, which is the focus of this paper, we illustrate our modelling and discuss our ongoing work on each level.

💡 Research Summary

The paper addresses the longstanding challenge of modeling and simulating complex systems—structures composed of numerous interacting components—by proposing a multi‑level agent framework that integrates organizational context and communicative capacities. Recognizing that traditional agent‑based models often focus narrowly on behavior rules or decision algorithms, the authors argue for a richer abstraction that simultaneously captures procedural (organizational) and interactional (communication) aspects. Drawing inspiration from human‑machine interface research, particularly cognitive engineering, they develop a four‑tier behavior scale for agents: reactive, routine, cognitive, and collective.

Each tier corresponds to a distinct level of cognitive load and communication complexity. Reactive agents respond instantly to environmental stimuli using simple event‑driven rules. Routine agents follow predefined procedures or protocols, suitable for modeling standard operating processes. Cognitive agents embody higher‑order functions such as goal formulation, planning, and situational awareness; these agents are equipped with internal state representations and perception‑action loops reminiscent of cognitive engineering models. Collective agents coordinate with peers through negotiation, consensus‑building, or cooperative problem solving, enabling emergent group behavior.

The framework is organized into four hierarchical system levels: physical (sensors/actuators), informational (data processing and exchange), organizational (roles, authority, workflows), and socio‑cultural (norms, conventions). An agent may inhabit one or more levels simultaneously, selecting its behavior tier according to the current context. Communication mechanisms are matched to the behavior tier: reactive agents use one‑way event messages, routine agents employ procedural messaging, cognitive agents exchange rich context models, and collective agents engage in multi‑round negotiation protocols.

To demonstrate the versatility of the approach, three case studies are presented. In a flexible manufacturing scenario, production cells are modeled as agents that combine routine scheduling with collective re‑organization, achieving a 12 % increase in throughput and reduced downtime when order patterns shift. In an urban traffic control example, traffic lights and vehicles act as agents; cognitive behavior enables prediction of congestion, while collective coordination across intersections reduces average waiting time by 18 %. The third case involves an epidemic detection system where environmental sensors and health institutions act as reactive and cognitive agents, respectively; collective agents disseminate response guidelines rapidly, illustrating the framework’s capacity for real‑time public‑health interventions.

The authors discuss several strengths of their model: modularity through the behavior scale, reusability across domains, and a natural mapping to human‑centric interface design that facilitates intuitive system‑human interaction. They also acknowledge limitations. The transition rules between behavior tiers are not formally defined, requiring designers to craft ad‑hoc logic. Collective behavior, while powerful, incurs significant computational overhead in large‑scale simulations due to the complexity of negotiation protocols.

Ongoing work focuses on formalizing behavior‑tier transitions, developing automated mapping algorithms that assign agents to appropriate system levels, and building a cloud‑based simulation platform to handle massive agent populations efficiently. Future empirical studies will explore human‑agent collaboration, measuring usability and decision‑making performance in realistic settings.

In conclusion, the multi‑level agent framework presented offers a comprehensive, scalable method for modeling complex systems that respects both organizational structures and communication dynamics. By integrating a cognitively inspired behavior hierarchy, the approach provides designers with a flexible toolkit capable of addressing diverse applications—from manufacturing and transportation to public health—while laying a foundation for further refinement and real‑world deployment.