Main Menu
About us
Project Description
Quantitative Results
Research Lines
Research Results
Impact on Society
Press room
Contact us
Secure Login
Events Calendar
« < April 2019 > »
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 1 2 3 4 5

Task 2 (UVEG): Development of a scalable multi-agent architecture for crowd simulation

Leader: Juan Manuel Orduña; Researchers: Guillermo Vigueras, Federico Silla (UPV), Carlos Reaño (UPV), Javier Nadal (UPV)

subTask 2.1 Design of a scalable multi-agent software architecture for crowds

subTask 2.2 Design of an efficient and scalable DVE system for crowd simulation

subTask 2.3 Design of a urban environment crowd system

subTask 2.4 Design of a crowd system for evacuation of public buildings

1. Brief Description of the Goals

Crowd simulations can be considered as virtual environment applications with two different goals. On the one hand, crowd simulations must focus on rendering visually plausible images of the environment, requiring a high computational cost. On the other hand, complex agents must have autonomous behaviours, greatly increasing the computational cost as well. Thus, some of the existing proposals can handle crowd dynamics and display populated interactive scenes, but they are not able to produce complex autonomous behaviours for their actors. On the contrary, some other proposals provide efficient and autonomous behaviours to crowd simulations, but they can only control a few hundreds of autonomous agents with different skills. That is, scalability has still to be solved in crowd simulations.

This research line focuses on developing a scalable multi-agent architecture that allows the simulation of large crowds of autonomous, intelligent agents while providing visually plausible images of the virtual world at good frame rates. In order to achieve this goal, our research addresses the system architecture (both hardware and software), as well as the implementation and social behaviour of agents. 

2. Scientific and Technical Developed Activities

We have developed a new system architecture specifically designed for crowd simulation published by Vigueras et al. in Journal of Network and Computing Applications, Vol. 32, No. 2, 2009. This architecture consists of implementing agents as process threads, in such a way that a single computer can hosts thousands of agents. Agents interact with other agents and with the virtual world through the Action Server, a central process in charge of controlling the virtual world. In order to properly scale up with the number of agents, the Action Server can be parallelized, published by Vigueras et al. in IEEE International Conference on Parallel Processing (ICPP), 2008. We have also developed different partitioning methods for efficiently distributing the action server, and we have shown that the shape of the regions in the partition has a major influence on the performance of the partitioning method, rather than the search method used, published by Vigueras et al. in 8th International Conference on Hybrid Intelligent Systems (HIS), 2008. Next, we have developed a scalable, distributed visualization system for crowd simulations, and we evaluated the integrated simulation and visualization systems, improving the performance and scalability of crowd simulations, published by Vigueras et al. in IEEE International Symposium on Parallel and Distributed Processing (IPDPS) 2010, published by Vigueras et al. in Appl. Soft Comput., vol. 10,  2010, by Vigueras et al. in Integrated Computer Aided Engineering, Vol. 18, Issue 4, 2011, by Vigueras et al. in The Journal of Supercomputing, Vol. 58, 2011 and by Vigueras et al. in Science of Computer Programming 2012. As well as, the PhD Thesis by Guillermo Vigueras was defended in 2011.

On other hand, we have developed a multiagent framework designed to animate socially intelligent agents, published by Grimaldo et al. in Intelligent Virtual Agents (IVA) 2007 and by Grimaldo et al. in Logic Journal of the IGPL, Vol. 19, 2011. Using this framework, we have implemented and evaluated a multi-modal decision making system for complex agents. In this way, we have evaluated different kinds of societies (egalitarian, elitist, etc) of complex agents, as well as several applications, published by Grimaldo et al. in International Joint Conference on Autonomous Agents and Multiagent Systems (AAMAS) 2008, by Grimaldo et al. in International Conference on Cyberworlds 2007, by Grimaldo et al. in Knowledge Eng. Review., Vol. 23, 2008, by Grimaldo et al. in International Conference on Practical Applications of Agents and Multiagent Systems (PAAMS) 2011 conference, by Grimaldo et al. in European Workshop in Multi-Agent Systems, EUMAS 2010 conference, by Grimaldo et al. in Multi-Agent-Based Simulation XII - International Workshop (MABS) 2011 and by Domínguez et al. in AAMAS 2012.

Additionally, we tuned a well-known multiagent platform for crowd simulation applications, published by Fernández-Bauset et al. in International Conference on Agents and Artificial Intelligence (ICAART) 2010 and by Fernández Bauset et al. in International Conference on Artificial Intelligence (AI) 2010.  

Publications: [Grim07], [Grim07b], [Grim07c], [Grim07d ], [Loza07 ], [Loza07b ], [Loza07c ],  [Grim08 ], [Grim08b ],  [Vigu08 ], [Vigu08b ], [Vigu08c ], [Loza09 ], [Vigu09 ], [Vigu09b ], [Grim09 ]

Projects funded by Public Calls: TIN2006-15516-C04-04, TIN2009-14475-C04-04, TIN2007-29664-E  by  national grants.

External collaborations Academia: Barcelona Supercomputing Center, (Spain)Rossaria Conte/Mario Paolucci, Alejandro Guerra Hernández.

External collaborations Industry: --

Company Agreements: --

PhD dissertations: Guillermo Vigueras González

Patents: --