Previous editions‎ > ‎PADABS 2013‎ > ‎

Program 2013

First Session: Welcome to PADABS and Tutorials

  • Welcome (Vittorio Scarano, Rosario De Chiara and Gennaro Cordasco)
  • Distributed Mason: A short tutorial (Vittorio Scarano, Università di Salerno, Italy)
  • The Rome Optimistic Simulator: A Tutorial (Alessandro Pellegrini, Università "La Sapienza" di Roma, Italy)
  • Tutorial on FLAME GPU (Daniela Romano, University of Sheffield (UK))
    • Slides (PDF)

Second Session: Parallel Agent-Based Simulations (Chair: Vittorio Scarano)

  • Ugo Erra and Giuseppe Caggianese. Parallel Hierarchical A* for Multi Agent-based Simulation on the GPU
    • Abstract: In this work, we describe a Parallel Hierarchical A* (PHA*) for path-finding in real-time using the Graphics Processor Units (GPUs). The technique aims to find potential paths for many hundreds of agents by building an abstraction graph of the input map in an off-line phase and then using this representation to speed up the path-finding during the on-line phase. The approach is appropriate in the case of scenarios based on grid maps and is independent on a specific topology. In addition, we propose also a strategy to obtain smooth paths during the search. We show that this approach fits well with the programming model of the GPUs, enabling searching for many hundreds of agents in parallel in real-time applications such as simulations. The paper describes this implementation using the Compute Unified Device Architecture programming environment, and demonstrates its advantages in terms of performance and quality of the paths founded comparing PHA* with a GPUs implementation of Real-Time Adaptive A* and the classic A* algorithm.
    • Slide (PDF)
  • Claudio Márquez, Eduardo César and Joan Sorribes. Agent Migration in HPC Systems using FLAME
    • Abstract: In HPC agent based applications, a large number of agents and complex interaction rules would likely cause workload imbalances that negatively affect the simulation time. In addition, the imbalance problem may vary during the application execution in accordance to the behavior of the agents. Consequently, a solution to this problem should be able to dynamically balance the load. A dynamic load balancing scheme could be based on migrating agents between processing units. In this paper, we propose a modification of the agent-based simulation framework FLAME that provides the automatic generation of the routines needed to dynamically migrate agents among different computational units. However, most agent-based simulation frameworks do not include routines for migrating agents. Moreover, we demonstrate their use in a simple load balancing scheme on a specific application.
    • Slides (PDF)
  • Gennaro Cordasco, Carmine Spagnuolo, Francesco Milone and Ada Mancuso. Communication Strategies in Distributed Agent-Based Simulations: the Experience with D-MASON
    • Abstract: Agent-based simulation models (ABMs) are a very powerful experimental tool of analysis, used in many scientific and technological communities of researchers, to assess and predict the dynamic unfolding of a series of events or processes, according to the imposition of certain conditions, given by the analyst. 
      The computing power usually represents a limit for such simulations and the traditional answer to the need for computing power is to invest in computer resources. 
      D-MASON is a framework for parallelize simulations developed on top of MASON toolkit. The goal of D-MASON is to exploit wasted computing power in a network of computers, eventually heterogeneous, as a research lab or a cluster. 
      In this paper we present a communication strategy using Publish/Subscribe paradigm through a layer based on the MPI Standard.
    • Slides (PDF)

Third Session: Parallel and Distributed Agent-Based Simulations (Chair: Gennaro Cordasco)

  • Guillaume Laville, Kamel Mazouzi, Christophe Lang, Nicolas Marilleau, Bénédicte Herrmann and Laurent Philippe. MCSMA: a toolkit to benefit from many-core architectures in agent based simulation
    • Abstract: Multi-agents models and simulations are used to describe complex systems in domains such as biological, geographical or ecological sciences. The increasing model complexity results in a growing need for computing resources and motivates the use of new architectures such as multi-cores and many-cores. Using them efficiently however remains a challenge in many models as it requires adaptations tailored to each program, using low-level code and libraries. In this paper we present MCSMA a generic toolkit allowing an efficient use of many-cores architectures through already defined data structures and kernels. This toolkit promotes few famous algorithms (diffusion, path-finding, population dynamics) which are ready to be used in an Agent Based Model. For other need, MCSMA is based on a flexible architecture and can easily be enriched by new algorithms thanks to development features. The use of the library is illustrated with two models and their performance analysis.
    • Slides (PDF)
  • Gennaro Cordasco, Rosario De Chiara, Fabio Fulgido and Mario Fiore Vitale. Supporting the exploratory nature of simulations in D-Mason
    • Abstract: Agent-Based Model (ABM) denotes a class of models which, simulating the behavior  of multiple agents (i.e., independent actions, interactions and adaptation), 
      aims to emulate and/or predict complex phenomena.  The ``emergence'' of such complex phenomena is often computation intensive and requires tools, libraries and frameworks that speed up and make easier to manage complex simulations. 
      In this paper we present new developments on D-Mason, that is a distributed version of Mason, a well-known and popular library for writing and running Agent-based simulations.  
      The new developments are: a) a tool that allow the parallel exploration of the behavior parameter space; b) an infrastructure that improves the management of distributed simulations in terms of easy deployment of new simulations, automatic update, visioning control and distributed logging.
    • Slides (PDF)
  • Fanny Boulaire, Mark Utting and Robin Drogemuller. Parallel ABM for electricity distribution grids: a case study
    • Abstract: This paper introduces a parallel implementation of an agent-based model applied to electricity distribution grids. A fine-grained shared memory parallel implementation is presented, detailing the way the agents are grouped and executed on a multi-threaded machine, as well as the way the model is built (in a composable manner) which is an aid to the parallelisation. Results gave medium level speedup of 2.6, but are hoped to be improved by combining this implementation to distributed or parallel ABM schedulers now available in the community. While domain-specific, this parallel algorithm can be applied to similarly structured ABMs (directed acyclic graphs).
    • Slides (PDF)

Fourth Session: Use Cases (Chair: Rosario De Chiara)

  • Mario Paolucci and Luca Vicidomini. A distributed simulation of roost-based selection for altruistic behavior in vampire bats
    • Abstract: Vampire bats exhibit a unique behaviour: they help unlucky conspecifics (that is, those who didn’t find enough blood to drink in their nightly hunt) by regurgitating in their mouth part of their semi-digested  meal. This altruistic act seems to happen in their group of reference even between individuals not genetically related. Inspired by this biological example, we have developed a simulation that reproduces the essential traits of the target phenomenon.
      Previous work on the same theme has shown how the roosting effect can cope with sensible mutation levels. Leveraging the features of D-Mason that allows to run massive simulations on a distributed environment, as well as performing a batch scheduling of the experiments, we explore the possibility to derive a relationship between mutation rate and roost size.
    • Slides (PDF)
  • Alessandro Pellegrini and Francesco Quaglia. A Study on the Parallelization of Terrain-Covering Ant Robots Simulations
    • Abstract: Agent-based simulation is used as a tool for supporting (time-critical) decision making in differentiated contexts. Hence, techniques for speeding up the execution of agent-based models, such as Parallel Discrete Event Simulation (PDES), are of great relevance/benefit. On the other hand, parallelism entails that the final output provided by the simulator should closely match the one provided by a traditional sequential run. This is not obvious given that, for performance and efficiency reasons, parallel simulation engines do not allow the evaluation of global predicates on the simulation model evolution with arbitrary time-granularity along the simulation time-axis. In this article we present a study on the effects of parallelization of agent-based simulations, focusing on complementary aspects such as performance and reliability of the provided simulation output.
      We target Terrain Covering Ant Robots (TCAR) simulations, which are useful in rescue scenarios to determine how many agents (i.e., robots) should be used to completely explore a certain terrain for possible victims within a given time.
    • Slides (PDF)
  • Michele Carillo, Nicola Lettieri, Domenico Parisi, Francesco Raia, Flavio Serrapica and Luca Vicidomini. Sociality, sanctions, damaging behaviors: a distributed implementation of an agent-based simulation model
    • Abstract: The explanatory and predictive power of social simulations is more and more connected with the development of models accounting for the complexity of real (inter-individual and intra-individual) social dynamics. From this perspective, a promising research path is complementing very simple models, more suitable to illuminate core dynamics of social phenomena, to increasingly more complex and empirically grounded simulations (big data-driven models, higher number of agents, more detailed and realistic description of cognitive and communication mechanisms underlying individual and group behaviors). The choice has two strictly intertwined effects: not only a different modeling approach, but also the need for more powerful tools. The paper presents a distributed implementation of an agent-based model exploring the interplay between damaging behaviors, sanctions and social mechanisms of learning and imitation, a topic investigated in many areas of social science from economics to legal science. Taking cue from a previous work based on a simple NetLogo simulation, the work shows how distributed solutions can help in developing more complex, wide and semantically rich models.
    • Slides (PDF)