Purpose This paper aims to propose a novel approach for optimizing coalition formation in distributed co-working networks. Traditional multi-agent collaboration optimization methods often struggle to cope with the combinatorial complexity and dynamic nature of real-world environments. The main goal is to address the NP-hard problem of optimal coalition formation while ensuring a balance between maximizing collaborative value and minimizing operational costs. Design/methodology/approach The framework combines Artificial Rabbit Optimization (ARO) with Particle Swarm Optimization (PSO). This integration leverages ARO's exploration capabilities with PSO's convergence properties. We introduce the Hybrid Convergence Theorem, providing theoretical guarantees that the proposed algorithm converges to near-optimal solutions within polynomial time bounds. Empirical validation is conducted using demographic and economic data from the Aosta Valley region, involving 100 agents distributed across 74 municipalities under realistic resource constraints. Findings Computational experiments show a 23.4% improvement in solution quality compared to genetic algorithms and an 18.9% reduction in computational time compared to exact methods. The framework effectively manages coalition sizes of up to 50 agents, while maintaining solution feasibility across multiple resource dimensions, including space, time, funding, and geographical constraints. Results confirm the effectiveness of the hybrid approach in balancing solution quality with reduced computational costs in complex and distributed environments. Originality/value This research introduces the first hybridization of ARO and PSO for coalition formation problems. Unlike traditional single-method approaches, the proposed framework unifies exploration and convergence mechanisms within a mathematically grounded structure. The introduction of the Hybrid Convergence Theorem provides new theoretical insights into hybrid metaheuristics. The study demonstrates how hybrid swarm intelligence methods can effectively tackle NP-hard optimization tasks under real-world constraints, offering a valuable tool for decision-makers in distributed networks and resource allocation scenarios.
Ciano et al. (Wed,) studied this question.