The Fire Seed Adaptive Diffusion and Retention Algorithm for Continuous Distributed Constraint Optimization Problems

Abstract

To address the challenges of proneness to local optima and population diversity decay in solving Continuous Distributed Constraint Optimization Problems (C-DCOPs), this paper proposes a novel meta-heuristic algorithm inspired by the phenomenon of spark propagation and persistence—the Fire Seed Adaptive Diffusion and Retention Algorithm (FSAD). Within this framework, a value of an individual in the population is regarded as a fire seed. By simulating the propagation, competition, and persistence processes of fire seeds, a four-stage coordinated distributed optimization framework is constructed. First, in the initialization phase, initial fire seeds are generated via controlled ran- dom sampling, and temperature parameters are configured based on problem-specific characteristics. Second, a multi-mode fire seed diffusion strategy is designed, integrating the targeted exploitation of elite fire seeds, the extensive exploration of random fire seeds, and the escape capability of reverse fire seeds. The core innovation lies in the introduction of a fire seed retention strategy, which adopts a simulated annealing-based probabilistic acceptance mechanism to ensure the persistence of superior fire seeds while dynamically maintaining population diversity. Additionally, the algorithm periodically implements a forced renewal mechanism for the worst-performing fire seeds, regularly eliminating low-vitality solutions to enhance global escape capability. Experimental results demonstrate that the FSAD algorithm outperforms traditional methods in terms of solution accuracy, convergence speed, and robustness, providing an efficient and reliable solution for C-DCOPs.