An Ant-Based Algorithm to Solve Distributed Constraint Optimization Problems

Authors: Ziyu Chen, Tengfei Wu, Yanchen Deng, Cheng Zhang

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical evaluation indicates that ACO DCOP is able to find solutions of equal or significantly higher quality than state-of-the-art DCOP algorithms. We empirically evaluate our proposed method with peer algorithms including DSA (p = 0.8), MGM2, DSAN, DGibbs, GDBA and the ant solver for DCSP (ACO DCSP for short) on two type of problems: random DCOPs and weighted graph coloring problems.
Researcher Affiliation Academia Ziyu Chen,1 Tengfei Wu,2 Yanchen Deng,3 Cheng Zhang1 Chongqing University 1 {chenziyu,bootan}@cqu.edu.cn, 2 wutengfei0404@163.com, 3 dyc941126@126.com
Pseudocode Yes Algorithm 1: Ant Colony Optimization
Open Source Code No The paper does not provide any specific link or explicit statement regarding the availability of open-source code for the described methodology.
Open Datasets No For random DCOPs, we set agent number to 70, domain size to 10 and uniformly select cost from [1, 100]. Each pair of agents is constrained independently with a probability p1 = 0.1 (for the sparse configuration) or p1 = 0.6 (for the dense configuration). We consider weighted graph coloring problems with 120 agents, 3 available colors for each agent and p1 = 0.05. The cost for each violation is uniformly selected from range 1 to 100.
Dataset Splits No The paper does not provide specific dataset split information (e.g., percentages, counts, or predefined splits) for training, validation, or testing.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory, or processor types) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes For random DCOPs, we set agent number to 70, domain size to 10 and uniformly select cost from [1, 100]. Each pair of agents is constrained independently with a probability p1 = 0.1 (for the sparse configuration) or p1 = 0.6 (for the dense configuration). For ant-based algorithms, we consider the number of ants K = 13 for the sparse random DCOPs, K = 20 for the dense random DCOPs and weighted graph coloring problems, set τ0 = 3 for ACO DCOP and τ0 = 10 for ACO DCSP. Moreover, we implement ACO DCSP as a pipelining version and set α = 2, β = 8, ρ = 0.02.