agent-based simulation - Sponsored Whitepaper

Agent based simulation has it's roots in theoretical mathematics since the 1940s. The concept has it's basis in the idea of large systems of interacting machine automata which can replicate, learn and reproduce. The early adoption of the technique in the simulation field has been evolutionary and has developed and matured significantly in recent years. Several influential studies in the field have occurred since the 1970s, leading to adoption into the field of general simulation as what we now know as agent based simulation. This paper discusses agent based simulation in the context of one such general purpose simulation system, flexsim. Flexsim is a visual object oriented simulation tool which has several core abilities of virtual reality, discrete event modeling, a built in rapid prototyping language and full integration with c++. Agent based simulation with the flexsim's o-o capabilities is discussed.

The origins of ABS date as far back as the 1940s when von neumann and stanislaw ulam developed the concepts of cellular automata - machines, which had a key characteristic of being able to reproduce themselves and act independently according to their defined behavior. In the 1970s john conway developed the influential ‘game of life’ - a game whereby cells on a grid are given simple rules to govern each turn of the game. The result was some surprisingly unpredictable behavior on the grid which is one of the first examples of system emergent behavior. Later in the 1980s craig reynolds devised a computer simulation named ‘biods’ where automata in 3d geometric space where given simple rules governing their movement with respect to each other. When the simulation was run, the resultant behavior of the system as a whole far surpassed the complexity of the rules governing each of the individual automata in the system. Thus the simulation became a compelling example of system emergent behavior. In the decades that followed, the concept of focusing on agent based simulation became something of a discipline in the world of simulation. In the present day, ABS is becoming integrated into general purpose simulation tools in combination with other modeling concepts such as process based modeling

ABS has no strict universal definition as yet, but is becoming a mature technology. It is essentially an object oriented approach to modeling. The behavior of the system to be modeled is automata-centric as opposed to process centric. An often cited feature of objects in an agent based simulation is their ability to learn during the simulation thus altering their behavior dynamically during a simulation run. A key feature of agents is their

awareness of their environment. An agent is designed to react dynamically to it's environment (the simulation model including other agents) and it's knowledge of it's environment is limited to a given scope so as to model it's localized observation of it's surroundings. A powerful feature of ABS is the phenomenon of emergent behavior. Emergent behavior is system wide behavior of the model as a whole which could not necessarily have been predicted in terms of the behavior of the individual agents.

Applications of agent based simulation include: