|
|
How
much advanced and efficient is the software your models is based on?, how
much friendly and interactive your model could be?
There is a growing number of modellers providing user-friendly on-line access to their models through Internet, and this number will rapidly increase in the future. A search into relevant webs have already identified a number of very interesting cases, covering different technologic solutions and scientific fields. Providing user-friendly and on-line access to models has a number of crutial advantages for both the modeller and the end-user (scientists and decision-makers). On the one hand, modellers could optimise the hardware and software tools supporting the model (since the model is actually installed in their own computer facilities) making them as productive and efficient as possible, and then provide on-line access to a potentially unlimited number of end-users. End-users could gain access to a large number of operational models and use them by themselves, removing the need to install each model (and all related databases) into their own computer system (which in most cases will be suboptimal handling advanced models, or simply incompatible). Outside the transport research programme there are a number of interesting experiences (from PACE-FORWARD high-level policy tool to Websites allowing travelers to explore the shortest or cheapest alternative route or transport service), but in general it is fair to say that the transport planning sector is using (nowadays) ICT development less than other sectors actually do. Next table provides a reference to the software architectures of transport models developed in the 4th FP.
The feasibility of the "on-line and friendly access to models" vision, requires solving two set of very different problems: -The interactivity gap (on-line communication between remote computer systems and models able to run on real-time, providing "just-in-time" outputs when feasible) -The user-friendliness gap (translation of users questions into models inputs and models outputs into meaningful answers) First assessment: Solving the interactivity gap: Getting access to remote models Basically, there are three options to get access to models (or to any software tool): -The Internet option (through Internet browsers) -The Intranet option (through Intranet Operative Systems) -The single computer option (by directly installing the model in the computer).
5.2 Solving the user-friendliness gap: Getting intelligent access to model The user-friendliness gap, from a software point of view, there is an increasing number of commercial applications providing useful tools to customise user-interfaces to end-users demands (both Internet and Intranet). Therefore, the key element is converting user-interface into actual knowledge tools able to help users to understand the interest and limits of the model, as well as to formulate legitimate questions and interpret models outputs. For very complex models, an Expert System (e.g. Bridges/DSS) maybe needed to carry on such task. ASSEMBLING is using three alternative software solutions to build up user-interfaces: -Internet interface connected to a Dynamic System model based on ASTRA-SCENES-SCENARIOS development. The actual model is located in the server of Karlsruhe University and runs in a Unix machine under Lynux OS, and the Internet interface establishes the link between the user and the model. The options available to the user are necessarily restricted only to those queries the model can answer on-time. -Internet interface based on a Java applet which includes an statistical formulation which simulates the actual model. The statistical formulation was calibrated based on previous runs of the model, which was originally programmed in Visual Basic using an spreadsheet as basis. -Intranet interface based on spreadsheets (EXCEL) which can be downloaded by free from ASSEMBLING website.
The first exploration of the subject has demonstrated, first, that there are many alternative software solutions to interface advanced models through Internet and Intranet; and second, that the development of any of these solutions requires to re-think the model as a software tool to be interlinked to others tools. In many cases it is convenient to re-programme the model or to calibrate a new statistical model based on the results provided by the actual model (after carrying on a sensitivity analysis, e.g.). In fact, the "interface" element between models and users has to be a "knowledge tool", able to translate users questions into models queries and models outputs into answers meaningful to the users. The communication between users and models must be interactive in order to allow for on-line validation of users intuitions and presumptions: it is by simulating the different impacts of alternative decisions, and evaluate the difference between users perceptions and model response, that users can be engaged in a self-reinforcing learning process.
Mcrit has carried out a first systematic search on Internet to find interesting samples of interactive models enjoying user-friendly interfaces. Based on these experiences, a systematic presentation has been developed (click to see it).
|
|
? |
||||||||||||||||||||||||||||||||||||
|
Objective
|
|||||||||||||||||||||||||||||||||||||||