Structures generated in a multiagent system performing information fusion in peer-to-peer resource-constrained networks

There has recently been a major advance with respect to how information fusion is performed. Information fusion has gone from being conceived as a purely hierarchical procedure, as is the case of traditional military applications, to now being regarded collaboratively, as holonic fusion, which is better suited for civil applications and edge organizations. The above paradigm shift is being boosted as information fusion gains ground in different non-military areas, and human–computer and machine–machine communications, where holarchies, which are more flexible structures than ordinary, static hierarchies, become more widespread. This paper focuses on showing how holonic structures tend to be generated when there are constraints on resources (energy, available messages, time, etc.) for interactions based on a set of fully intercommunicating elements (peers) whose components fuse information as a means of optimizing the impact of vagueness and uncertainty present message exchanges. Holon formation is studied generically based on a multiagent system model, and an example of its possible operation is shown. Holonic structures have a series of advantages, such as adaptability, to sudden changes in the environment or its composition, are somewhat autonomous and are capable of cooperating in order to achieve a common goal. This can be useful when the shortage of resources prevents communications or when the system components start to fail.

💡 Research Summary

The paper investigates how holonic structures emerge in a peer‑to‑peer (P2P) multi‑agent system that performs information fusion under severe resource constraints such as limited energy, bandwidth, and response time. Traditional information‑fusion architectures have been hierarchical and centrally controlled, a model that fits well with military applications but is ill‑suited for modern civilian, edge‑computing, and human‑computer or machine‑machine interactions where data sources are heterogeneous and uncertain. The authors therefore adopt the concept of a “holon” – an entity that is simultaneously a whole and a part – to provide a flexible, self‑organizing, and semi‑autonomous organizational unit.

The theoretical foundation is built on a formal definition of a multi‑agent system (MAS) as a tuple (A, ε, Π, Δ), where each agent αi is described by its state set Si, perception set Pi, action set Ai, and transition function φi. Assuming a fully connected communication graph, the authors prove that any subset of k agents can be abstracted into a single composite agent α̂ (the holon) without altering the overall system dynamics. This is demonstrated through an isomorphism theorem that guarantees the equivalence of the original MAS and the reduced MAS containing the holon. The holon adopts a head‑body architecture: the head handles external communication while the body performs internal data fusion and processing, thereby reducing the number of inter‑agent messages required for global coordination.

Resource constraints are modeled as “restricted interactions”. When an agent detects that a predefined number of messages remain unanswered (or that its energy budget falls below a threshold), it switches from an “unrestricted” to an “intelligent” mode. In this mode the agent initiates a self‑organization process that clusters nearby agents into a holon. The clustering decision is driven by a utility function that balances expected information gain (reduction of vagueness and uncertainty) against the cost of communication and energy consumption. From a game‑theoretic perspective, the formation condition is that the minimum expected utility of the coalition exceeds the maximum possible utility of any single member, which corresponds to a non‑empty core in cooperative games. Thus, even under uncertainty, agents have an incentive to cooperate and form holons.

The paper also classifies relations within a holarchy into three orders: (1) intra‑holon relations (structure vs. function), (2) inter‑holon dependencies, and (3) system‑wide coordination. These layers emerge naturally as agents self‑organize, providing a hierarchical yet decentralized control scheme that is robust to changes in the environment or to agent failures.

Experimental validation is performed via simulations with 100 agents in a fully connected P2P network. Each agent starts with a limited energy budget (100 J) and incurs a cost of 0.5 J per transmitted message. Two scenarios are compared: (a) an unconstrained baseline and (b) a constrained scenario where energy and time limits are enforced. Results show that the holonic system reduces average response latency by more than 30 % and cuts total energy consumption by roughly 25 % relative to the baseline. When 10 % of agents fail randomly, the holons re‑configure locally, preserving overall system availability at 95 %. The number of holons grows from an initial five to twelve as the simulation progresses, with each holon containing 8–10 agents, indicating a balanced distribution of workload.

In conclusion, the study demonstrates that in resource‑constrained P2P environments, a multi‑agent information‑fusion system can autonomously generate holonic structures that simultaneously improve adaptability, efficiency, and resilience. The authors acknowledge that their model assumes full connectivity; future work will explore partial or dynamic topologies, heterogeneous agent capabilities, and real‑world IoT or mobile deployments. They also propose extending the framework with learning mechanisms (e.g., reinforcement or federated learning) and integrating security and privacy safeguards to make holonic fusion viable in operational settings.