What drives autonomous vehicles?

Imagine calling a taxi using one of the  cab aggregating apps and you find a ‘ghost’ driver. If you think it is a page out of a science fiction, think again as Autonomous Vehicles (AV) or self-driving vehicles are already in operation in some cities of the world on a small scale. For instance, nuTonomy, a Singapore-based startup, became the world’s first company to test a self-driving taxi service in Singapore on August 25, 2016.

Just a few weeks later, on September 14, 2016, Uber launched its first self-driving taxi fleet in Pittsburgh, Pennsylvania, USA. You will hear more such news in the days to come as AVs are expected to replace the conventional driver-driven vehicles by 2020 or sooner. A study by Business Insider, a news website, estimates about 10 million self-driving cars to be on roads by 2020.

The market for AV is big and presently has automobile giants like BMW,
Volkswagen and Tesla racing against each other to transfer the driving control from humans to computers. Google’s ambitious Self Driving Car project, developed in
collaboration with Stanford University, was a game changer and won the 2006 US Defense Advanced Research Projects Agency (DARPA) Grand challenge. While companies like Tesla are developing a computer-based driver assistance system to help drivers avoid accidents, others are relying on the concept of a fully automated car.

Algorithm-based
The quest for AVs started with the need to fix the least reliable part of the car — the driver. A look at the statistics can emphasise the need for this — over 1.4 lakh people are killed in road accidents every year in India and about 93% of these accidents are caused due to human error. Computers, unlike humans, are near perfect drivers who respond quickly.

Various experiments have been conducted on automating cars since the 1920s and the first truly autonomous car was realised in the 1980s at the Carnegie Mellon University’s Navlab, USA. Research picked up pace in the recent decade with the advent of improved machine learning techniques, better computing hardware and software, and more accurate sensor technologies.

However, there are reports of collisions of such cars that have been termed as either a software failure or a misunderstanding. Google has confirmed that there have been 12 collisions of its self-driving cars as of 2015 with just one resulting in injuries to passengers in the car. On 14 February, 2016, a Google self-driving car struck a bus when it attempted to avoid sandbags blocking its path. Such crashes have been termed as ‘learning experience’.

So, what actually drives these cars? The brain of the AVs is the algorithms that are programmed to function in a specific way based on Artificial Intelligence and
Machine Learning. These algorithms, much like our brain, react based on the surrounding conditions. A self-driving car uses a GPS to know its location, a RADAR to detect obstacles, a laser ranging system to map the three-dimensional surrounding and a camera to identify traffic lights and signs, recognise pedestrians and monitor other vehicles on the road. There are also sensors that track weather and road conditions. All the data collected from the cameras, GPS, RADARs and sensors are fed to an on-board computer which manoeuvres the car to its destination, avoiding people and other vehicles on the way. However, for rare events like a plastic bag blowing down the road, the computer has to rely on the ‘knowledge’ developed through test drives. This is where the learning comes into play. Based on the inputs and the learning history, the computer decides the course of action and thus, making the ride safe and reliable.

The advantages of AVs do not stop at that. AVs can then be sold as a service, just like cabs, and no one would ever want to own a car! What’s more is that these vehicles will also be eco-friendly with a five to 15% reduction in fuel consumption and a comparable reduction in CO2 emissions. However, not everything is rosy at the moment with self-driving vehicles.

A cause for concern?
Though there is a push by several governments and industries for these vehicles, there are some ethical, legal and security issues that need to be addressed. Situations where a car has to decide between the lives of passengers and the person who is in front are still unresolved and need to be debated widely. Legal issues concerning accidents involving AVs as who should be blamed by law need to be charted out. There is also a widespread skepticism about hackers taking over the control while someone is on board.
As a ray of hope, some of steps in the right direction is being taken with the US Federal government drafting guidelines for automated vehicles. At this point, not everyone might know how AVs might change their life, but there is a sense of excitement. But can a computer compete against humans? Only time can tell!

(The author is with Gubbi Labs, a Bengaluru-based research collective)