Moumita Choudhury

Distributed Constraint Optimization Problems (DCOPs) have been widely used to coordinate interactions in cooperative multi-agent systems. The Functional DCOP (F-DCOP) model extends the traditional model to continuous variables. The existing F-DCOP algorithms experience huge computation and communication overhead. This paper applies continuous non-linear optimization methods on the Cooperative Constraint Approximation (CoCoA) algorithm and empirically shows the algorithm provides high-quality solutions at smaller cost.

@article{sarker2020c,
  title={C-CoCoA: A Continuous Cooperative Constraint Approximation
         Algorithm to Solve Functional DCOPs},
  author={Sarker, Amit and Arif, Abdullahil Baki and Choudhury, Moumita
          and Khan, Md. Mosaddek},
  journal={Proceedings of AAMAS 2020},
  year={2020}
}

We combine DCOPs and Functional DCOPs into the Mixed Integer Functional DCOP (MIF-DCOP) framework, and propose Distributed Parallel Simulated Annealing (DPSA), where agents cooperatively learn the optimal parameter configuration while solving the problem. DPSA produces solutions of significantly better quality than state-of-the-art non-exact algorithms.

@inproceedings{mahmud2020learning,
  title={Learning Optimal Temperature Region for Solving Mixed Integer
         Functional DCOPs},
  author={Mahmud, Saaduddin and Khan, Md. Mosaddek and Choudhury, Moumita
          and Tran-Thanh, Long and Jennings, Nicholas R},
  booktitle={Proceedings of IJCAI},
  pages={268--275}, year={2020}
}

We present Anytime Evolutionary DCOP (AED), a novel population-based algorithm that uses evolutionary optimization to solve DCOPs. Agents cooperatively construct random solutions and improve them through a mechanism considering optimistic local benefit approximations. We prove AED is anytime and show it outperforms state-of-the-art DCOP algorithms in solution quality.

@inproceedings{mahmud2020aed,
  title={AED: An Anytime Evolutionary DCOP},
  author={Mahmud, Saaduddin and Choudhury, Moumita and Khan, Md. Mosaddek
          and Tran-Thanh, Long and Jennings, Nicholas R},
  booktitle={Proceedings of AAMAS},
  pages={825--833}, year={2020}
}

We propose Particle Swarm based F-DCOP (PFD), a new algorithm for Functional DCOPs inspired by Particle Swarm Optimization. PFD devises a distributed method that significantly reduces computation and memory requirements. We prove PFD is an anytime algorithm, and empirical results show it outperforms state-of-the-art approaches in solution quality and overhead.

@inproceedings{choudhury2020particle,
  title={A particle swarm based algorithm for functional distributed
         constraint optimization problems},
  author={Choudhury, Moumita and Mahmud, Saaduddin and Khan, Md Mosaddek},
  booktitle={Proceedings of AAAI},
  volume={34}, number={05},
  pages={7111--7118}, year={2020}
}