New insights into how London taxi drivers plan their routes could inform the development of better navigation tools, reports a study led by UCL, University of York and Champalimaud Foundation researchers.
Famous for having to pass a test on "The Knowledge" by learning the layouts of more than 26,000 streets, London's cabbies are expert navigators that researchers are particularly interested in learning from.
According to the new Proceedings of the National Academy of Sciences (PNAS) study, London taxi drivers rationally plan each route by prioritising the most challenging areas first and filling in the rest of the route around these tricky points. This is different to satnavs, which calculate every possible route until they reach the destination.
Current computational models to understand human planning systems are challenging to apply to the 'real world' or at large scale, and so researchers measured the thinking time of London taxi drivers while they planned journeys to various destinations in the capital city.
The researchers believe that the flexible planning strategies used by London taxi drivers could benefit the development of future AI navigation technologies.
Previous studies have shown the uniqueness of the London taxi driver's brain; a famous series of papers led by the late Professor Eleanor Maguire at UCL found they have a larger posterior hippocampus region (a part of the brain involved in learning and memory) than the average person, with their brain changing in volume as a result of their cab-driving experience.
First author Dr Pablo Fernandez Velasco, visiting research fellow in the Spiers Lab at UCL and British Academy postdoctoral fellow at the University of York, said: "London is incredibly complex, so planning a journey in a car 'off the top of your head' and at speed is a remarkable achievement.
"If taxi drivers were planning routes sequentially, as most people do, street-by-street, we would expect their response times to change significantly depending on how far they are along the route.
"Instead, they look at the entire network of streets, prioritising the most important junctions on the route first, using theoretical metrics to determine what is important. This is a highly efficient way of planning, and it is the first time that we are able to study it in action."
For the study, the research team asked 43 taxi drivers to plan routes between two designated sites in London and call out the routes step-by-step - similar to what they are asked to do on "The Knowledge" examination to obtain their license to drive a taxi in London.
The researchers showed that taxi drivers use their cognitive resources in a much more efficient way than current technology and argue that learning about expert human planners can help with AI development in a number of ways.
Joint senior author Dr Dan McNamee (Champalimaud Foundation, Portugal) said: "The development of future AI navigation technologies could benefit from the flexible planning strategies of humans, particularly when there are a lot of environmental features and dynamics that have to be taken into account.
"Another way to enhance these technologies would be to integrate the information about human experts into AI algorithms designed to collaborate with humans. This is a very important point, because if we want to optimise how an AI algorithm interacts with a human, the algorithm has to 'know' how the human thinks."
Lead author Professor Hugo Spiers (UCL Psychology & Language Sciences) added: "This study certainly confirms what other studies have found - the London taxi driver's brain is incredibly efficient, and its larger volume is put to good use in making sense of a highly complex city like London."
The ongoing Taxi Brains project, led by Professor Spiers's spatial cognition lab, also involves taking MRI brain scans of cabbies, particularly their hippocampus, in order to contribute to Alzheimer's disease research, as navigational skills and the hippocampus are impaired by the disease.
The research is supported by the British Academy, the EPSRC UK, and Ordnance Survey.
