Artificial Intelligence Automates Driving Tests

Artificial intelligence (AI) is a mega trend that will affect most processes and people in the coming years. Here’s another example.

Microsoft Research has developed a smartphone-based driving test system that uses the power of AI to make a fair analysis of a learner driver’s ability before issuing a licence. The technology underlying the driving test is called HAMS, short for Harnessing Automobiles for Safety. It was originally conceived as a system to monitor drivers and their driving skills, with the aim of improving road safety.

Microsoft recently announced that the system has been deployed at the Regional Transport Office in Dehradun, India in the state of Uttarakhand and is ready for wider adoption in India and beyond.

The system uses a smartphone affixed to a car’s windshield. Through the smartphone’s front and rear cameras and other sensors, HAMS simultaneously monitors the driver — for instance, their gaze — and the road scene, checking, for example, the distance to the vehicle in front. It employs advanced AI models, which the Microsoft team has developed for efficient and robust operation.

The HAMS technology deployed in these driving tests has been customized to include capabilities such as precise tracking of a car’s trajectory during designated test manoeuvres, for instance, parallel parking or navigating a roundabout. This tracking enables HAMS to determine precisely, at any instance, whether the driver stopped in the middle of a manoeuvre for longer than is permitted or tried to correct the manoeuvre by rolling forward and backward alternately more times than allowed.

HAMS, running on a smartphone and on an edge server onsite at the testing track, produces a detailed report shortly after the driver completes the exercises. The driving test ensures transparency, and drivers have the option of viewing a video recording of their test drive, to know if the system treated them fairly.

Microsoft and other major tech companies have been using India as a test bed to build solutions for other markets around the globe. For example, the company has developed tools to help farmers in the country grow their crop yields, and worked with hospitals to prevent avoidable blindness. In 2018, Microsoft partnered with Apollo Hospitals to create an AI-powered API customized to predict the risk of heart diseases in India. Several popular apps from other providers, such as YouTube Go and Google Station, also started as India-only services.

Automation and AI are slowly making their way into driver testing around the world, but the technology still needs deployment of extensive infrastructure like pole-mounted video cameras along test tracks. Adoption of HAMS can bring down the cost of automation and improve test coverage by including a view within the vehicle.

The success of HAMS in Dehradun bodes well for making the technology available at a lower price and in a more accessible way. This will help its wider adoption in India and other countries, and is another example of how AI will be involved in so many aspects of our lives.

That said, Microsoft has limited involvement in the development of self-driving cars, so it doesn’t have the same depth of understanding of the sheer complexity of road conditions as others. For example, it was reported last week that the Uber self-driving car that killed a pedestrian in Arizona in March 2018 was not designed to detect pedestrians jaywalking. Many countries will want to put AI itself through a large number of driving tests before letting it loose on human drivers.