Science in Singapore

"HALF an hour is plenty," Babbage thought to himself at precisely 8.30am as he jumped in a cab to take him from the expatriate enclave around Holland Road to a gathering at the National University of Singapore (NUS) merely 5.7 km (3.5 miles) away. How naive. Singapore's roads have grown more congested since the 1990s, when he lived here as a teenager. Soon after hitting the main local thoroughfare the taxi was stuck, its driver shamelessly trying to shed responsibility by asking for advice as to what route to follow in these sorry circumstances. Hazy childhood memories suggested one that was circuitous but ought to take less time. Whether or not it did shall forever remain a niggling counterfactual. Either way, the taxi eventually pulled up at the NUS at 9.07 am, seven minutes late. 

Ironically, Babbage would have probably made it on time had the ideas presented there already come to pass. The Future Urban Mobility (FM) workshop, organised by the Singapore-MIT alliance for Research and Technology (SMART), was all about how to make city living, and especially city driving, a less pitiful experience. Although FM has only been up and running in earnest for six months, the project has already yielded a plethora of noteworthy results.

Of particular relevance to your correspondent were those arrived at by Daniela Rus, of the Massachusetts Institute of Technology. Her SMART team has been trying catch a glimpse of Singaporeans' driving patterns. To do this, they have analysed a month's worth of data from ComfortDelGro, which runs Singapore's largest taxi fleet of some 15,200 vehicles. These included precise GPS coordinates and speeds for all the company's cabs, as well as customer pick-up and drop-off points, for a total of 512m data points (or 33GB).

Of course it is far from obvious that cabbies' habits reflect those of other drivers. For one thing, it is unlikely that many residents wake up at 6am with an irrepressible urge to jump in their cars and head straight for the airport, as a clear spike in the taxi data would indicate. However, such anomalies need not pose a problem. So long as they are systematic, as in the airport case, they can be corrected for. (Pollsters do this all the time to make their samples more representative of the general population.)

To check whether they are systematic, Jay Aslam, a member of Dr Rus's SMART posse from Northeastern University in America, is now cross-referencing overall traffic flows for a number of intersections with the taxi figures for those same spots at the same times. He is hoping that the trove of taxi data might reveal some interesting congestion patterns emerging on Singapore's roads. 

However, to Dr Rus and Dr Aslam, knowing how people drive is merely the first step in an effort to make driving more efficient. Together with Sejoon Lim, a graduate student at MIT, they are busily developing tools which use the traffic flow information to chart the best route for a person trying to get from A to B. It turns out that this often depends on how much time he has to spare. While there may be just one path which offers any odds of reaching B in a jiffy, several more options are typically available to a less time-pressed traveller. For instance, the quick path may offer a better than even chance of making it in 15 minutes, with all alternative routes in effect guaranteeing that he will not reach B on time. Where time is less scarce, that same route might offer a 90% probability of success, say, but one of the other paths may well do even better. In other words, a traveller prepared to settle for a lower payout (ie, a longer trip) can limit the risk of failure (a late arrival); a high-risk strategy may bring high rewards (a speedier arrival), but typically at the cost of higher uncertainty about the outcome. The wily driver ought always to plump for the path which minimises the chances of arriving late.

Your correspondent could certainly have done with these insights earlier that morning. Then again, he could simply have left home a bit sooner.