Large-scale agent-based modelling and simulations

Agent-based modelling has become an indispensable tool for analysing the behavior of complex socio-technical systems. We perform basic and applied research as well as technology development in agent-based modeling and simulation.

Objectives

  • Large-scale complex simulation toolkits -- We research and develop agent-based simulation frameworks that allows rapidly building specific models of a wide variety of real-world systems. In contrast to existing agent-based simulation platforms, our toolkits support complex, structured physical environments (as opposed to simple grid and network spaces) and deliberative cognitive agents (as opposed to simple reactive agents). Our frameworks have been scaled up to support up to millions of agents in a single model.
  • Combining agent-based and discrete-event simulation -- Traditionally, agent-based simulations use a fixed time-stepped execution model. We explore how the agent-based modelling paradigm can be combined with the discrete-event simulation approaches that allow better efficiency than time-stepped model. We view the agent-based model as a layer on top of a discrete event processing core which aims as an implementation layer for handling simulation environment state updates.
  • Simulation development methodologies -- We aim to provide a simple yet flexible methodologies for building agent-based simulation models. Currently, our focus is primarily on providing guidelines and formalisms for analysis and specification of agent-based models in terms of agents, behaviors, environments, sensors and actuators.
  • Simulation-based experimental methodologies -- Simulations are increasingly more used as virtual testbeds for evaluating and validating multi-agent control mechanisms. We systematically study this problem and are developing a methodology for mixed-reality testbeds which combine real-world assets with their digital simulation counterparts in order to accelerate the development of complex human-agent-robot applications. 

Approach

The centre has vast experience with building a variety of agent-based models in a number of applications domains. The experience forms a rich knowledge base for generalization of the specific models and development trails into a more general, robust frameworks. Understanding that there is always a fundamental trade-off between the universality of frameworks and the added value such frameworks provide out-of-the box for solving specific problems, we strive to stay at the sweet spot of universality and specificity. Although much of our technology and know-how is application-agnostic, we have a strong application track in modelling air traffic management, tactical urban operations, maritime traffic and multi-modal transport systems.

Results

Throughout the history of the centre, we have developed a range of simulation frameworks and tools. As of today, the following are under active development (click through for more details):

  • Alite -- software toolkit helping with particular implementation steps during construction of multi-agent simulations and multi-agent systems in general.
  • AgentFly -- agent-based model of air traffic and aerial vehicles (both manned and unmanned).
  • AgentPolis -- fully agent-based toolkit for modeling multi-modal transportation systems (based on Alite) .
  • AgentC -- agent-based model of maritime traffic and activity in piracy-affected waters.
  • AgentScout -- agent-based model of tactical urban operations.
  • AgentDrive -- agent-based model of road and highway vehicular traffic.
  • AgentCrowd -- agent-based model of public gatherings We have also published a series of papers on employing mixed-reality testbeds for the development of multi-agent system.

People

Michal Jakob, Antonin Komenda, David Sislak, Premysl Volf, Ondrej Vanek, Ondrej Hrstka, Zbynek Moler