The Capacity Restraint Vine: a Powerful Framework for Modelling Individual Travellers Dynamic Decision Making in a Network at Micro Level
HOFFMANN N, Halcrow Fox, UK
Favourable developments in Electronic Computing have in recent years seen the emergence of a variety of Road and Pedestrian Micro Simulation Program Systems, many of which make use of traditional non-dynamic assignment methods, suggesting a need for a mor
Favourable developments in Electronic Computing have in recent years seen the emergence of a variety of Road and Pedestrian Micro Simulation Program Systems, many of which make use of traditional non-dynamic assignment methods, suggesting a need for a more dynamic approach.
Almost ten years ago, not long after the King's Cross station fire in 1987, where 31 people died, the author's Company was asked to provide "dynamic" routing within London Underground's Station Congestion Model (SCM) for the purpose of modelling station emergency evacuation scenarios. "Dynamic" meaning: create new routes on demand and reassigning flows to the new routes, in accordance with the assignment principles of SATURN 1. A modified version of SATURN 2 had until then been used for providing all static routes to the detailed simulation program SCM. At the time it was recognised that a more dynamic approach was needed, at least when modelling station evacuations, and ideally in the general use of SCM as well.
At the time, SCM could only deal with complete routes, and some lessons had already been leamed from the use of the SATURN 2 routing to satisfy SCM's insatiable appetite for routes: the demand for routes often exceeded the supply of computer memory. In addition the distribution of routes would often show an imbalance in the use of parallel facilities, e.g. banks of escalators and stairs, causing unnecessary congestion at times:, a function of the compromise of not always being able to provide all necessary routes.
The concept of the Capacity Restrained Vine (CRV) was considered, at this stage, the only tenable way of creating routes, given the hardware constraint at the time, and PEDROUTE3 3 was born. Later creating routes had to be abandoned in favour of creating link progressio n probabilities, as a means of splitting rate type assignment,due to more and more complex models running out of routes. Eventually, even progression probabilities could at times be exhausted.
About seven years ago, when developing PAXPORT 4, on behalf of British Airports Authorities (BAA), the CRV concept was developed further. It was found that modelling the complexity of an airport necessitated modelling individuals, since travellers' would be moving from a to b, b to c and onwards, spending available free time in between visiting compulsory facilities, such as security and check-in. Demand flows are therefor often difficult to define.
Association for European Transport