Wednesday, July 18, 2007

Simulation and Modelling - New Opportunities?

In recent months we've seen a major step function made in the financial services industry as grid computing is applied in a widespread fashion for the purposes of providing scalable resources for pricing, risk and modelling activities. The proposition is that the provision of a large scale distributed computing infrastructure is beginning to power new models of business to new heights of performance. Perhaps the most compelling example is in the use of Algorithmic trading models, where automated trading is taking the place of human intervention, based on the design of models for watching and reacting to market developments. The New Scientist had an excellent overview of this in its May 30 Issue . The key question is where else do we have opportunities for such innovation, changing the way in which work is done?

Well there are several areas of industry and government where simulation and modelling can break the mould, introducing innovation in the way in which work gets done. One such example is the airplane design industry where new composite materials, such as those introduced in the Boeing 787 Dreamliner, have the potential to revolutionise the construction of airframes. Similarly, the design of cars of all sorts in the automotive industry are being accelerated through the design of platform technologies and the rapid ability to generate variants to address specific market niches.

A recent meeting with the Transport Modelling industry, the potential for grid computing to revolutionise the planning of transport systems became apparent to those present. There are opportunities to speed up the single thread software applications currently in use through multiple execution in virtualised environments; there is potential to capitalise on simulation techniques used in the microsimulation models now promoted by companies such as SIAS who provide simulations of human behaviour as part of their modelling solution. This allows for the range of human responses to the same stimuli which we encounter in our journeys and thus a richer modelling result. There are folks who need to deliver traffic activity data into the scheduling and traffic adviser industries upon which businesses and individuals increasingly rely to make their journeys more efficient and/or arrive at their expected destination!

Finally, there are the big-picture transport planners from the Department for Transport's national road and freight models to Transport for London, National Rail network and London Underground, all of whom are building infrastructures to support our lives in the 21st Century. Much of the work they are doing is focused on one particular aspect; and many of the approaches adopted have their roots in historical approaches, scaled up for the challenges of today. The consequence of this are limited modelling processes which are lengthy in execution, think of days, operating on relatively narrow, independent data sets without recourse to the enrichment of data that is possible in the Google Earth/Mash-Up worlds of today.

For those of you in the science community you may recognise this pattern, its where we were 5 years ago before the Grid and e-Science community got to grips with the equivalent problems in Astronomy; Biology; Chemistry and other fields of exploration. There is some promising evidence of the potential for federating data, a project is underway at Cambridge University; and there are some interesting new ideas percolating in the area of simulation and modelling which can: a) show more graphically the patterns of activity in our transport network; b) provide better tools for reasoning about planning choices and c) provide a more compelling infrastructure for implementation. It's a brave new world out there. How can we use it to change the game to dramatically improve the way we do our work?