Improving human supervision: Microsoft and India’s driving license tests
India is a complex country, a place where simple issues in other countries, such as obtaining a driver’s license, can be made difficult due to issues ranging from all kinds of irregularities, including corruption.
According to some studies, 59% of drivers on the road have never passed or taken their driving test, a burden that can be avoided by paying an official or, sometimes, obtaining a false license (around 30% of the total) or by sending another driver to take the exam. In India, this issue is a huge source of concern: 20% of the total 1.35 million deaths resulting from car accidents each year worldwide occur in India, around a quarter of a million, or one every four minutes.
Driving tests, which take place on specialist circuits, has traditionally been carried out in the applicant’s vehicle with an examiner, which raised numerous questions to do with subjectivity rather than the driver’s ability.
Now, Microsoft has developed HAMS (Harnessing AutoMobile Safety), an app that replaces the driving instructor or examiner with a smartphone attached to the windshield of the vehicle and that controls, in addition to confirming the identity of the driver, the test, using both the front and rear cameras of the device and all details of the movement of the vehicle (using GPS and accelerometer). This provides a full picture of the exam, taking into account maneuvering and even whether drivers have checked mirrors.
Using a device instead of an examiner means greater productivity at driving test centers — which usually carry out around 40 exams per day, up to 130 in the case of Delhi — in addition to introducing greater transparency, because there is now a recording of the exam that can be revised in case of doubt.
Microsoft has come up with a solution that runs on cheap and widely available hardware, that will introduce traceability and reliability, as well as objectivity and transparency for a procedure that undoubtedly needs it. What’s more, it is highly valuable in terms of automation and application development based on issues such as computer vision and machine learning.
(En español, aquí)