Mesoscopic Models: An Alternative to Massive Microsi

Microsimulation is what many people (consultants, clients and stakeholders) now expect when they talk about traffic modelling.

It does an excellent job of showing clients how their roads and improvement schemes may work. The visual representation of roads and vehicles is very powerful, and can allow everyone, from client to layman, to understand what the implications of a road scheme may be. But have we started to ask too much of microsimulation?

As we have seen the benefits of microsimulation, there has been a temptation to use it for larger and larger modelled areas. In most cases we have had to choose from either ‘large’ (in microsimulation terms) models with questionable worth due to greatly increased build and run times, or multiple smaller, inconsistent models that always just miss the first question asked of “what happens just outside the network?”.

Over the last few years, software developers have been working hard to provide us with solutions to these problems and there are new trends emerging; in particular mesoscopic and hybrid models. This progress matches interest from clients and we are really starting to see a shift towards authorities building larger multi-tier level models that can answer both local and strategic questions, in one package.

How does this differ from other simulation models? Macro is perfect for understanding large regional areas, whereas Micro can show us the detail of interaction between road users (even to the Nano level of how pedestrian and vehicle movements are affected when a crossing is added). However, the two are not really compatible. The very nature of Macro means it has little constraint on capacity, which often means the flows on the roads are far greater than a (capacity constrained) Micro model can cope with. This leaves many problems for any modeller or client trying to make use of both Macro and Micro level modelling, consistently.

Could mesoscopic simulation be the solution? Meso gives us most of the speed of Macro, as simulation of vehicle behaviour is reduced to moving a vehicle from one road section to the next, without simulating the vehicle in detail within the section. In Meso, individual vehicles are still simulated, we still use the same lanes and signals as in Micro but less information about precise vehicle movements in the road section is generated.

However, if hybrid Meso-Micro models are developed, we get the best of all worlds with the individual vehicle from Meso travelling through the Micro sub areas, whilst exchanging information dynamically across the wider network. We work in detail in the Micro areas, but see the effect of changes in the wider Meso area without the need to calibrate to Micro standards, and with simulation run time benefit. And of course as we are dealing with individual vehicles, we keep the advantages of being able to model dynamic route choice and to gather data about journey time variability.

This futureproof method promotes consistency between all of the modelling layers as it can be done in one model, keeping a clear audit trail and consistent results.

Alastair Kitson, Senior Traffic Engineer akitson@systra.com