Speeding Up Incomplete GDL-based Algorithms for Multi-agent Optimization with Dense Local Utilities
IJCAI, pp. 31-38, 2020.
EI
Weibo:
Abstract:
Incomplete GDL-based algorithms including Max-sum and its variants are important methods for multi-agent optimization. However, they face a significant scalability challenge as the computational overhead grows exponentially with respect to the arity of each utility function. Generic Domain Pruning (GDP) technique reduces the computational...More
Code:
Data:
Introduction
- Distributed Constraint Optimization Problems (DCOPs) [Modi et al, 2005; Fioretto et al, 2018] are a fundamental model for multi-agent optimization and coordination, in which agents cooperatively find assignments to optimize a global objective.
- Max-sum and its variants [Farinelli et al, 2008; Rogers et al, 2011; Zivan et al, 2017; Chen et al, 2018] are popular incomplete algorithms built upon the Generalized Distributive Law (GDL) [Aji and McEliece, 2000] and have been applied to many real-world domains due to their ability of directly handling n-ary constraints.
- The computational effort grows exponentially with respect to the arity of each utility function, which prohibits Max-sum from scaling up to large systems
Highlights
- Distributed Constraint Optimization Problems (DCOPs) [Modi et al, 2005; Fioretto et al, 2018] are a fundamental model for multi-agent optimization and coordination, in which agents cooperatively find assignments to optimize a global objective
- Max-sum and its variants [Farinelli et al, 2008; Rogers et al, 2011; Zivan et al, 2017; Chen et al, 2018] are popular incomplete algorithms built upon the Generalized Distributive Law (GDL) [Aji and McEliece, 2000] and have been applied to many real-world domains due to their ability of directly handling n-ary constraints
- We empirically evaluate the performance of Generic Domain Pruning (GDP), FDSP and our proposed algorithms when accelerating Max-sum on both random DCOPs and NetRad systems
- We demonstrate that the presence of dense utility functions or utility ties would deteriorate the performance of GDP
- We present a discretization mechanism which offers a tradeoff between the reconstruction overhead and pruning efficiency. We theoretically show their correctness and superiorities over GDP
- We plan to extend our work to cope with changing utility functions in a dynamic environment
Methods
- The quality of the lower bound is the key of effective pruning, especially when local utilities are dense.
- Alg.1 presents the sketch of GD2P.
- When computing a response message for a variable node xi ∈ xj, it searches for the highest utility for each assignment vi ∈ Di by exhausting the sorted entries whose local utility is no less than the running lower bound lb in a sequential order.
- The lower bound is updated whenever a higher utility is found
Results
- The authors generate high arity and low arity factor graphs by uniformly selecting an arity for each function node from [2,5] and [5,8], respectively.
- FDSP outperforms domain pruning variants, it is still dominated by ART-GD2P.
- This is due to the fact that with the sorted AND/OR trees the proposed ART-GD2P can find a high-quality lower bound quickly despite of highly-structured utility functions.
- ART-GD2P with a large t reduces the runtimes in both preprocessing phase and pruning phase, which demonstrates the necessity of the discretization mechanism
Conclusion
- The authors demonstrate that the presence of dense utility functions or utility ties would deteriorate the performance of GDP.
- The authors present a discretization mechanism which offers a tradeoff between the reconstruction overhead and pruning efficiency.
- The authors theoretically show their correctness and superiorities over GDP.
- The authors plan to extend the work to cope with changing utility functions in a dynamic environment
Summary
Introduction:
Distributed Constraint Optimization Problems (DCOPs) [Modi et al, 2005; Fioretto et al, 2018] are a fundamental model for multi-agent optimization and coordination, in which agents cooperatively find assignments to optimize a global objective.- Max-sum and its variants [Farinelli et al, 2008; Rogers et al, 2011; Zivan et al, 2017; Chen et al, 2018] are popular incomplete algorithms built upon the Generalized Distributive Law (GDL) [Aji and McEliece, 2000] and have been applied to many real-world domains due to their ability of directly handling n-ary constraints.
- The computational effort grows exponentially with respect to the arity of each utility function, which prohibits Max-sum from scaling up to large systems
Objectives:
The authors aim to cope with dense local utilities from the perspectives of both bound quality and search space organization and develop more efficient sorting-based acceleration algorithms.- The authors aim to develop more efficient sorting-based acceleration algorithms for incomplete GDL-based algorithms by alleviating the negative effect of dense local utilities
Methods:
The quality of the lower bound is the key of effective pruning, especially when local utilities are dense.- Alg.1 presents the sketch of GD2P.
- When computing a response message for a variable node xi ∈ xj, it searches for the highest utility for each assignment vi ∈ Di by exhausting the sorted entries whose local utility is no less than the running lower bound lb in a sequential order.
- The lower bound is updated whenever a higher utility is found
Results:
The authors generate high arity and low arity factor graphs by uniformly selecting an arity for each function node from [2,5] and [5,8], respectively.- FDSP outperforms domain pruning variants, it is still dominated by ART-GD2P.
- This is due to the fact that with the sorted AND/OR trees the proposed ART-GD2P can find a high-quality lower bound quickly despite of highly-structured utility functions.
- ART-GD2P with a large t reduces the runtimes in both preprocessing phase and pruning phase, which demonstrates the necessity of the discretization mechanism
Conclusion:
The authors demonstrate that the presence of dense utility functions or utility ties would deteriorate the performance of GDP.- The authors present a discretization mechanism which offers a tradeoff between the reconstruction overhead and pruning efficiency.
- The authors theoretically show their correctness and superiorities over GDP.
- The authors plan to extend the work to cope with changing utility functions in a dynamic environment
Funding
- This research is supported by Singapore National Research Foundation projects AISG-RP2019-0013, NSOE-TSS201901, and NTU
Reference
- [Aji and McEliece, 2000] Srinivas M Aji and Robert J McEliece. The generalized distributive law. IEEE Transactions on Information Theory, 46(2):325–343, 2000.
- [Chen et al., 2018] Ziyu Chen, Yanchen Deng, Tengfei Wu, and Zhongshi He. A class of iterative refined Max-sum algorithms via non-consecutive value propagation strategies. Autonomous Agents and Multi-Agent Systems, 32(6):822– 860, 2018.
- [Chen et al., 2019] Ziyu Chen, Xingqiong Jiang, Yanchen Deng, Dingding Chen, and Zhongshi He. A generic approach to accelerating belief propagation based incomplete algorithms for DCOPs via a branch-and-bound technique. In AAAI, pages 6038–6045, 2019.
- [Cohen et al., 2020] Liel Cohen, Rotem Galiki, and Roie Zivan. Governing convergence of Max-sum on DCOPs through damping and splitting. Artificial Intelligence, 279:103212, 2020.
- [Farinelli et al., 2008] Alessandro Farinelli, Alex Rogers, Adrian Petcu, and Nicholas R Jennings. Decentralised coordination of low-power embedded devices using the Maxsum algorithm. In AAMAS, pages 639–646, 2008.
- [Fioretto et al., 2017] Ferdinando Fioretto, William Yeoh, Enrico Pontelli, Ye Ma, and Satishkumar J Ranade. A distributed constraint optimization (DCOP) approach to the economic dispatch with demand response. In AAMAS, pages 999–1007, 2017.
- [Fioretto et al., 2018] Ferdinando Fioretto, Enrico Pontelli, and William Yeoh. Distributed constraint optimization problems and applications: A survey. Journal of Artificial Intelligence Research, 61:623–698, 2018.
- [Hirayama and Yokoo, 1997] Katsutoshi Hirayama and Makoto Yokoo. Distributed partial constraint satisfaction problem. In CP, pages 222–236, 1997.
- [Hirayama et al., 2019] K Hirayama, K Miyake, T Shiotani, and T Okimoto. DSSA+: Distributed collision avoidance algorithm in an environment where both course and speed changes are allowed. International Journal on Marine Navigation and Safety of Sea Transportation, 13(1):117– 123, 2019.
- [Hoang et al., 2018] Khoi D Hoang, Ferdinando Fioretto, William Yeoh, Enrico Pontelli, and Roie Zivan. A large neighboring search schema for multi-agent optimization. In CP, pages 688–706, 2018.
- [Khan et al., 2018] Md Mosaddek Khan, Long Tran-Thanh, and Nicholas R Jennings. A generic domain pruning technique for GDL-based DCOP algorithms in cooperative multi-agent systems. In AAMAS, pages 1595–1603, 2018.
- [Kim and Lesser, 2013] Yoonheui Kim and Victor Lesser. Improved Max-sum algorithm for DCOP with n-ary constraints. In AAMAS, pages 191–198, 2013.
- [Kim et al., 2011] Yoonheui Kim, Michael Krainin, and Victor Lesser. Effective variants of the Max-sum algorithm for radar coordination and scheduling. In WI/IAT, pages 357–364, 2011.
- [Kschischang et al., 2001] Frank R Kschischang, Brendan J Frey, Hans-Andrea Loeliger, et al. Factor graphs and the sum-product algorithm. IEEE Transactions on Information Theory, 47(2):498–519, 2001.
- [Li et al., 2016] Shijie Li, Rudy R. Negenborn, and Gabriel Lodewijks. Distributed constraint optimization for addressing vessel rotation planning problems. Engineering Applications of Artificial Intelligence, 48:159–172, 2016.
- [Litov and Meisels, 2017] Omer Litov and Amnon Meisels. Forward bounding on pseudo-trees for DCOPs and ADCOPs. Artificial Intelligence, 252:83–99, 2017.
- [Macarthur et al., 2011] Kathryn S. Macarthur, Ruben Stranders, Sarvapali D. Ramchurn, and Nicholas R. Jennings. A distributed anytime algorithm for dynamic task allocation in multi-agent systems. In AAAI, pages 701–706, 2011.
- [Maheswaran et al., 2004] Rajiv T Maheswaran, Jonathan P Pearce, and Milind Tambe. Distributed algorithms for DCOP: A graphical-game-based approach. In ISCA PDCS, pages 432–439, 2004.
- [Modi et al., 2005] Pragnesh Jay Modi, Wei-Min Shen, Milind Tambe, and Makoto Yokoo. Adopt: Asynchronous distributed constraint optimization with quality guarantees. Artificial Intelligence, 161(1-2):149–180, 2005.
- [Monteiro et al., 2012] Tania L Monteiro, Guy Pujolle, Marcelo E Pellenz, Manoel C Penna, and Richard Demo Souza. A multi-agent approach to optimal channel assignment in WLANs. In WCNC, pages 2637–2642, 2012.
- [Netzer et al., 2012] Arnon Netzer, Alon Grubshtein, and Amnon Meisels. Concurrent forward bounding for distributed constraint optimization problems. Artificial Intelligence, 193:186–216, 2012.
- [Nguyen et al., 2019] Duc Thien Nguyen, William Yeoh, Hoong Chuin Lau, and Roie Zivan. Distributed Gibbs: A linear-space sampling-based DCOP algorithm. Journal of Artificial Intelligence Research, 64:705–748, 2019.
- [Okamoto et al., 2016] Steven Okamoto, Roie Zivan, and Aviv Nahon. Distributed breakout: Beyond satisfaction. In IJCAI, pages 447–453, 2016.
- [Ottens et al., 2017] Brammert Ottens, Christos Dimitrakakis, and Boi Faltings. DUCT: An upper confidence bound approach to distributed constraint optimization problems. ACM Transactions on Intelligent Systems and Technology, 8(5):69:1–69:27, 2017.
- [Pepyne et al., 2007] D Pepyne, D Westbrook, B Philips, E Lyons, M Zink, and J Kurose. A design for distributed collaborative adaptive sensing of the atmosphere. Technical report, University of Massachusetts, 2007.
- [Petcu and Faltings, 2005] Adrian Petcu and Boi Faltings. A scalable method for multiagent constraint optimization. In IJCAI, pages 266–271, 2005.
- [Rogers et al., 2011] Alex Rogers, Alessandro Farinelli, Ruben Stranders, and Nicholas R Jennings. Bounded approximate decentralised coordination via the Max-sum algorithm. Artificial Intelligence, 175(2):730–759, 2011.
- [Stranders et al., 2009] Ruben Stranders, Alessandro Farinelli, Alex Rogers, and Nicholas R. Jennings. Decentralised coordination of mobile sensors using the Max-sum algorithm. In IJCAI, pages 299–304, 2009.
- [Weiss and Freeman, 2001] Yair Weiss and William T Freeman. On the optimality of solutions of the Maxproduct belief-propagation algorithm in arbitrary graphs. IEEE Transactions on Information Theory, 47(2):736– 744, 2001.
- [Yeoh et al., 2010] William Yeoh, Ariel Felner, and Sven Koenig. BnB-ADOPT: An asynchronous branch-andbound DCOP algorithm. Journal of Artificial Intelligence Research, 38:85–133, 2010.
- [Zhang et al., 2005] Weixiong Zhang, Guandong Wang, Zhao Xing, and Lars Wittenburg. Distributed stochastic search and distributed breakout: Properties, comparison and applications to constraint optimization problems in sensor networks. Artificial Intelligence, 161(1-2):55–87, 2005.
- [Zivan et al., 2017] Roie Zivan, Tomer Parash, Liel Cohen, Hilla Peled, and Steven Okamoto. Balancing exploration and exploitation in incomplete Min/Max-sum inference for distributed constraint optimization. Autonomous Agents and Multi-Agent Systems, 31(5):1165–1207, 2017.
Full Text
Tags
Comments