Aimsun 6 The integrated transport modelling software

Mature and reliable models

When it comes to the models themselves, practitioners can feel comfortable that the algorithms in Aimsun are world-class - which means open to public scrutiny. We choose and implement models that can be, and are, subjected to exacting independent tests. Before introducing new models we carry out extensive assessments of all available alternatives with only one criterion in mind: choosing what consistently works best in the real world. Finally, as part of TSS’s commitment to openness, we not only supply every detail of our algorithms in the User’s Manual but also allow our users to substitute their own models, through the Software Development Kit (SDK). Static assignment

Aimsun features a multi-class user equilibrium assignment model that supports several important tasks:

• assigning traffic onto a network at macroscopic level under user equilibrium
• adjusting OD matrices to improve their quality after importing detection or other real-world data
• performing “traversals” i.e. computing automatically a consistent OD matrix for any sub-model of a larger model
• evaluating the quality of an existing detection layout
• recommending additional detector locations to improve coverage and minimise error

Microscopic simulation

The microscopic simulator in Aimsun features mature, best in class car-following and lane-changing models. In independent benchmarking tests Aimsun routinely comes top against fierce commercial and academic competition.

Calibrating an Aimsun model is an important step and requires method and rigour - and a proportion of a project’s time - but it is a well-structured process free from obscurity.

All parameters have an intuitive, real-world meaning and are thoroughly documented. There are enough parameters to do a good job of matching real-world data but not too many to turn the calibration process into a vicious cycle. Perturbations of inputs bring about robust and predictable system responses and the software includes both a calibration assistant and a network checker.

From a performance point of view, the micro-simulator in Aimsun is in a league of its own: our micro-simulator runs a model of Singapore with 10,580 intersections and 4,483 km of lanes at 2-3 times faster than real time. A model of Barcelona city centre with a typical 1-hour demand micro-simulates in 1 minute (60 times faster than real time). These figures debunk the myth that micro-simulation inherently does not scale.

Mesoscopic simulation

The new mesocopic simulator in Aimsun 6 provides an additional option to practitioners who wish to model dynamic aspects of very large networks.

The new model works with individual vehicles and uses similar, but somewhat simplified car-following and lane-changing models which removes most of the calibration burden when compared to a micro-simulator. By modelling each section of a network as having a free-flowing and queuing part, the mesoscopic approach concentrates on key “events” and is able to advance the simulation clock between events. This simplifies the computational work and improves performance by orders of magnitude.

Most importantly, the mesoscopic simulator in Aimsun bases its models on tried and tested components used in the micro-simulator. The result, as measured on our benchmark models, is a performance gain of about 450% with an agreement in outputs between the two models that reaches an R2 value of 0.94.

Dynamic Traffic Assignment

Apart from the ability to perform static equilibrium assignments, Aimsun features two different Dynamic Traffic Assignment schemes, which can be used in either a mesoscopic or a microscopic simulation:

• a discrete choice model in which drivers take the cheapest path subject to a statistical distribution which can be one of three available by default (logit, c-logit, or proportional) or be defined by the user through Aimsun’s powerful function editor
• a Dynamic User Equilibrium scheme in which drivers select routes in accordance with a generalised equilibrium principle

During a simulation, users may designate previously derived assignments as historical routes; implement different traffic management strategies; define classes of vehicles that have different route choice behaviour; and restrict lanes for the exclusive use of buses or HOVs.