design metabolic metaheuristics

This work proposes a metabolic multi-agent optimization framework grounded in endogenous resource cycling, addressing the limited intrinsic adaptability of traditional metaheuristic algorithms that rely on external scheduling mechanisms. The framework employs a shared metabolic controller to dynamically regulate agent lifecycles, role transitions, and resource allocation, thereby establishing a lightweight, self-consistent closed-loop resource system. It unifies energy budgeting, normalized reward signals, and continuous role adaptation, enabling effective handling of both continuous and discrete optimization problems. Empirical evaluations on benchmark functions—including Sphere and Rastrigin—as well as synthetic and TSPLIB traveling salesman problem instances demonstrate the method’s cross-domain stability and computational efficiency, highlighting its compact architecture and general-purpose adaptive capabilities.

This study investigates whether the resource cycling mechanism in Metabolic Multi-Agent Optimizers (MMAO) possesses framework-level explanatory power beyond metaphorical interpretation. By constructing an abstract state model that strips away domain-specific details while preserving the core resource accounting mechanism, the work leverages abstract modeling, dynamical analysis, and stability theory under mild bounded gain-and-expenditure assumptions to reveal, for the first time, the endogenous regulatory dynamics of MMAO at the architectural level. The analysis establishes the non-negativity and boundedness of key resource variables—such as private energy and public budget—and identifies three universal behavioral modes: contraction, reinvestment, and search reallocation. The generality of these modes is validated through both continuous and discrete instantiations, clearly distinguishing behaviors inherent to the metabolic feedback loop from those contingent on implementation specifics.

This work proposes a metabolic multi-agent optimization framework that addresses the limitations of traditional metaheuristics, which rely on fixed population sizes and manually tuned search parameters and thus lack intrinsic adaptability. The framework introduces, for the first time, an endogenous metabolic resource allocation mechanism into optimization algorithms: agents carry energy, role states, and memory, enabling self-calibrating collaborative search through private–public metabolic resource cycles. In continuous domains, it employs energy-regulated symmetric zeroth-order probing and role-interpolated movement; in discrete domains, it integrates structure-aware guided perturbations with energy-weighted edge reuse. Experimental results demonstrate that this lightweight, low-parameter approach effectively achieves heterogeneous, self-adaptive cooperative search behavior without external parameter tuning, as validated on CEC2017 (10D/30D) and TSPLIB benchmarks.

This study systematically evaluates the closed-loop resource allocation mechanism of the Metabolic Multi-Agent Optimizer (MMAO) under a unified and strict budget constraint, demonstrating its effectiveness across both continuous and discrete optimization problems. Leveraging benchmark suites from CEC2017, TSPLIB, and OR-Library, the authors conduct large-scale empirical assessments against strong baselines—including PSO-lite, ES-lite, and iterative greedy 2-opt—to establish MMAO as the first cross-domain adaptive framework of its kind. Through trajectory-level diagnostics and ablation studies, the work reveals the robustness of MMAO’s endogenous resource reallocation capability. Results show that MMAO significantly outperforms baseline methods on both problem types, while ablated variants exhibit performance nearly matching the full model, confirming its ability to dynamically and efficiently allocate computational resources even under stringent budget limitations.

This study addresses the vulnerability of local structures in dynamic optimization environments by proposing an extension to the Metabolic Multi-Agent Optimizer (MMAO) that operates without external adaptation modules. The approach leverages MMAO’s endogenous metabolic mechanisms—comprising private energy, public budget, role drift, success feedback, and lifecycle turnover—and maps them onto non-stationary environments to enable autonomous dynamic adaptation. Evaluated on dynamic continuous optimization benchmarks (shifted Sphere, Ackley, and Rastrigin functions), the method achieves an average offline error of 28.07 across 216 trials, significantly outperforming the standard MMAO and other dynamic baselines. Notably, it demonstrates superior robustness and post-perturbation recovery on Sphere and Rastrigin functions, providing the first empirical validation that MMAO’s intrinsic metabolic cycle can independently drive efficient dynamic optimization behavior.

本文提出一种基于图结构的元启发式方法Mycelial Search，通过活性尖端、社区加权流、自适应线缆塑性和基于锚点的注入机制解决连续优化问题。

This study investigates whether intermittent search or Lévy walks are more advantageous in finite, depletable environments. To address this question, we construct a resource-consuming environment on a two-dimensional grid and employ genetic algorithms to drive the free evolution of movement strategies, thereby eliminating assumptions regarding predefined power-law distributions. The dynamical characteristics of the resulting search trajectories are analyzed by integrating Lévy dust modeling with second- and fourth-order displacement moment fitting. Our findings reveal that, under resource-constrained conditions, intermittent search consistently outperforms strict Lévy motion, achieving model fitting coefficients exceeding R² > 0.99. This work provides a theoretical foundation for designing exploration strategies employed by autonomous systems operating within environments characterized by finite resources.

This work proposes a “fluid agents” framework that introduces agent generativity into multi-agent reinforcement learning, thereby relaxing the conventional assumption of a fixed population size. Traditional approaches struggle to handle dynamic group structures where agents may join, leave, or even be generated by other agents—common in real-world scenarios. By integrating game-theoretic modeling with tailored reinforcement learning algorithms, the framework enables emergent team formation and adaptation. Evaluated in extended Predator-Prey and Level-Based Foraging environments, the approach demonstrates the ability to train agent systems that autonomously adjust team size in response to task complexity. Experimental results show that this adaptive capability yields superior collaborative strategies and task performance compared to settings with fixed populations.

研究通过基于规则的逆生物合成方法，使用Qwen2.5-7B策略选择扩展分子，提高了在给定扩展次数下的解题率。

本文提出OptiMAS，通过统一的ReAct基础架构和双轨记忆机制，解决多智能体系统自动优化中范围与稳定性之间的权衡问题。