The core issue is that general-purpose driving is an extremely complex problem to solve. To understand the challenge, we must break it down into three parts.- Perception, prediction and planning.
22 Feb 2020 | 6571 Views | By Prakash Chandra Pathak, BetterPlace
In Feb 2019, Elon Musk announced that Tesla cars would be fully autonomous by the end of 2020. People would be able to just summon the car from parking and enjoy a snooze in the back seat while the car drives them to their destination. Tesla will also launch an ...
This is an Autocar Pro Plus article. Subscribe to continue reading.
In Feb 2019, Elon Musk announced that Tesla cars would be fully autonomous by the end of 2020. People would be able to just summon the car from parking and enjoy a snooze in the back seat while the car drives them to their destination. Tesla will also launch an app-based ride hailing service.
Alphabet’s (the parent company of Google) self-driving company, Waymo, also launched a self-driving taxi service called Waymo One in Phoenix, Arizona in 2018. The company had started testing their self-driving cars in San Francisco, in 2009 under the Alphabet name, but later rebranded itself as Waymo. Waymo’s self-driving cars have covered 10 million+ miles on public roads so far.
There are also many other companies with promising self-driving technology, with Uber and nVidia among the leaders. Traditional automobile companies such as Toyota, Mercedes, and General Motors are also getting into this space.
What’s taking so long?
Many ask why it’s taking so long for self-driving technology to become widely available. The core issue is that general purpose driving is an extremely complex problem to solve. To understand the challenge, we must break it down into three parts.
The first part is that the self-driving car needs to fully understand its location and its surrounding. This is called perception. Perception comes very naturally to humans, but a self-driving car needs to understand its surroundings using cameras, radars, ultrasonic sensors and lidar (light detection and ranging). The current level of technology in self-driving cars does a pretty good job of understanding the data captured by these devices and makes a 3D model of its surroundings. However, just knowing the surroundings is not enough; the car also needs to know how the surroundings are going to change.
This brings us to the second part of the problem - prediction. In order to navigate traffic safely, a self-driving car must predict how traffic will move and change in the immediate future. Self-driving cars predict the movements of other vehicles (and pedestrians) by interpreting emitted turn signals, traffic lights etc.
The third problem is in planning. Cars must plan their actions based on predictions of changes in their surroundings. This, however, is far more complicated than it sounds. At any intersection, for instance, a self-driving car can predict that another automobile is going to make a turn based on its turn signal. But, when the self-driving car itself makes a move, human drivers would react in some way.. So acting upon its prediction modifies the self-driving car’s surroundings in such a way that invalidates the prediction that it just made. Human drivers negotiate in a way which is not easy to program into a computer.
Its difficult to determine when exactly true self-driving cars would be ready for use by the general public. It can be said with reasonable certainty that such cars would more or less be ready for public use in less congested areas of most American cities within 5 years.
Fast forward to 2025
Imagine that you are visiting a client in Santa Clara, California. You exit San Francisco airport and grab your phone to book an Uber. Your ride arrives, but it doesn’t have a driver. An automated voice greets you, “Welcome Mr Sharma. How was your flight ?” in an Indian accent. Uber used your name and home location to personalize the voice and message that the car greets you with.
You are a bit skeptical, but you see everyone else boarding these cars, so you shrug and get in. The 28 minute ride to your client’s office was mostly uneventful and, after the first 5 minutes, you didn't even notice that you didn't have a human driver behind the wheel. The car stops at the entrance of your client’s office and you get out. The car says “Have a wonderful day, Mr Sharma!”. You respond, “Thank you!”, knowing fully well that you just thanked a machine. Your first experience traveling in a self-driving was actually quite good.
Early stages of self-driving cars
During the early stages of public availability, self-driving cars will likely have few users. During this time, most people will probably request a self-driving car only when a human driver is not available, such as during peak traffic hours or late at night. Also, because self-driving cars need a lot of extra hardware, their production costs will be very high compared to standard automobiles, which would be prohibitive for wide-spread use. However, economies of scale and Moore’s Law will reduce the per km rates slowly.
Also people will eventually become comfortable with self-drive technology and then it will start to pick up. The first people affected will be truck drivers as highway driving is something that self-driving cars will master first. Falling prices of self-driving cars would cause job losses for truck drivers (and eventually cab drivers as well).
Expected job losses in the era of autonomous driving
Job losses in the drivers’ space alone are expected to be huge - it’s difficult to put a number to it today but can roughly be more than 50%. This is likely to trigger protests from their Unions, knowing fully well that progress can only be delayed but not stopped, this technology would eventually become the defacto standard.
Some may argue that job losses that will happen in the US and other developed countries will not have any impact in India. This is because our country has a long history of ignoring western technology and business models because we think that they would never work in India. Google Maps, for instance, is a technology that most of us thought would never work in India owing to our unstructured city architecture, traffic flows, and other annoyances. Before Google Maps, we used to rely on strangers on the street for directions. Now Google Maps has become an integral part of any driving experience here as well.
Same thing happened with car rental services like Hertz and Enterprise. It was predicted by analysts that such premium services won’t pick up in our country because trusting that people would return the car was difficult. However, Zoomcar and DrivEzy have made the system work here. So, like other critical technologies, self-driving cars will also adapt and eventually become a viable mode of transport in India too.
How would we respond as Indians?
Initial days of self driving cars would be pretty much similar to those in the USA. People will look at them with curiosity and suspicion. First few riders would have mixed experience. We may expect a few accidents or long traffic jams due to mistakes made by self driving cars. To tackle such issues, our Government should ready a policy draft before self driving cars become a reality here. Simultaneously, these cars would learn and if their makers constantly update the AI and software, the performance is likely to improve.
So what kind of jobs drivers would be doing in this future?
First thing is that new people would stop getting into the driving profession way before existing drivers would start losing jobs. Many of the people who work as drivers right now, also work as farmers in farming season. Currently, India is facing a shortage of farm labour. Self driving cars would push a big chunk of current drivers back to farming.
Another big segment of existing drivers would get into service industry working as a personal driver, escort and care takers for their customers. Many people rely on their drivers for far more than just driving. Drivers serve by opening doors, carrying luggage, guiding their customers through city finding places of business or suggesting some nice restaurant. These qualities would make drivers more valuable, enabling them to earn far more than what they currently do.
Yet another segment of drivers would upgrade their skills and become manager/caretaker of self driving cars. When thousands of self driving cars would be doing ten’s of trips on a daily basis, they would definitely need a fleet manager, who would be responsible to track these self driving cars. Initially, each car will have a caretaker who would just sit behind the wheel and intervene infrequently. But once self driving cars would become more reliable, one caretaker would be able to manage 8-10 cars. His duties would include watching status of all parameters of car like fueling, battery charge, general maintenance and driving the car in manual mode and bring to service station if required.
Automation need not be feared
It’s premature to say whether the comfort of self driving cars would make all the existing drivers job less, or would create new more fulfilling and higher paying jobs for drivers. In the past, we have seen automation displacing workers but also creating better jobs elsewhere.
Also technology has overall created a better life for the entire society. Case in point are multiple utility apps like cab hailing, grocery delivery, errands running and food delivery which has taken away a massive amount of negative labour from our lives and has freed up our time for more constructive work.
We hope that technology would keep enhancing human life in this new age of Artificial Intelligence automation.