At the upcoming AAG conference in New York, I’ll be presenting a recent prototype that links agent-based simulation with current traffic flow models.
The basic premise is that any cognitive decision associated with movement around cities should be modelled at the level of the individual. However, it is not always necessary that all movement be represented individually. Doing so potentially wastes limited computational power, especially important where modelling many complex agents.
Instead, my new simulation utilises traffic flow modelling to constrain the movement of individual agents. Individuals choose where they move individually, but physical movement itself is modelled collectively. The higher the traffic flow on a single route, the slower each agent on that route will travel. This approach is more efficient and allows a much larger scale of complex agent-based simulation.
I’ll provide more detail at AAG next Sunday, but the basic result is as above.
The simulation demonstrates traffic flows across central London. There are 30000 agents of varying behavioural characteristics moving around this space. Their movement decisions impact on the state of the network.
KEY: The redder colours represent high traffic saturation aka queues and congestion, the blues and greens represent quiet or free flowing traffic conditions.