RUSH hour in Nairobi can be a nightmare. 鈥淢ost motorists don鈥檛 follow traffic rules and small inconveniences like a minor traffic accident or even a sudden downpour can cause delays of up to an hour,鈥� says John Kimani, a small business owner in the Kenyan capital.
A text message service launched this month should help. Twende Twende (Swahili for 鈥淟et鈥檚 go, let鈥檚 go鈥�) provides motorists with updates on traffic conditions and advises on alternative routes. The system was developed by Nairobi-based IBM Research Africa, the firm鈥檚 first research laboratory in Africa.
Finding a reliable source of traffic information was the first challenge. Cities in developed countries have vast networks of traffic cameras, but Nairobi has just 36, operated by local internet service provider, Access Kenya. IBM Research Africa鈥檚 chief scientist, Uyi Stewart, used this camera data coupled with an image-enhancing algorithm to identify individual vehicles and their velocity. Next, he borrowed an inference algorithm from IBM鈥檚 Tokyo research lab and set about fine-tuning it to work in Nairobi. 鈥淒riving patterns are very different in Africa,鈥� says Stewart.
Advertisement
The adapted algorithm extrapolates traffic conditions on the 98 per cent of Nairobi鈥檚 streets not covered by cameras. 鈥淚f we have two isolated parallel streets with a single connecting street, and we know the ingress and egress numbers for cars on each of the parallel streets, we can estimate how many cars are taking a street connecting them,鈥� he says.
This algorithm is the basis of Twende Twende. To use it, motorists send a free text message that says where they are and where they want to go. In return, they receive a message with traffic conditions or a route recommendation.
聯Motorists send the service a message that says where they are and they receive a traffic update聰
Carlo Ratti of the SENSEable City Laboratory at the Massachusetts Institute of Technology believes that crowd-sourced travel apps will be more effective in the long run. But these require smartphones that come with built in GPS, and more than 88 per cent of Kenyans still use basic handsets without GPS capability.
Stewart wants as many people as possible to use Twende Twende, as this helps the algorithm to improve its recommendations over time. 鈥淭he more they use it, the more the system learns, the better it gets,鈥� he says.
