Everyday people worldwide spend so much time to move from one place to another place. Shouldn’t it be easier and safer to go wherever we want to go ? Since the last decade researchers, scientists and tech companies are putting in so much effort to develop fully self driving technology, and testing it on real city streets.
Self driving cars aren’t a science fiction anymore. Companies like Toyota and Ford have billions of dollars in Research and Development pouring into this technology. Services like Uber and Lyft who currently pay human drivers would soon deploy entire fleet of self driving cars. In few years we would see self driving cars being sold to regular consumers. But how do they work ?
How do they work ?
When we humans are in the driver’s seat, we are observing our environment by receiving an input of our surroundings and simultaneously processing it in order to make a decision of which way to turn the steering wheel. A self driving car is usually outfitted with GPS Unit, an inertial navigation system and a range of sensors. It uses the positional information from GPS and navigation system to localize itself and the sensor data to build an internal map of its environment. Once it has its position in its internal map of the world, it can then use that map to find the optimal path to its destination that avoids any kind of obstacles. That’s a very high level description of how self driving cars work.
Why Self Driving Cars at all ?
There are couple of reasons for their existence, lets us take a look at each of them one at a time.
1. Economics:
Utilization: Cars are seriously underutilized, a car is utilized 4% of its time, 96% it is parked in parking lot. A car is amongst the most important investment people invest in and still is underutilized. So we have a very expensive equipment available most of the time which is under utilized.
Cost: If we look at services which offers mobility on demand, such as Uber, Lyft, etc and analyse the cost per kilometer, the driver is 50% of the cost. If you remove the driver out of the loop and if you move to electric fuel as well, the cost per kilometer reduces significantly.
2. Add On Services :
There is also another dimension to it, that pulls tech companies into it. Since Car has passengers in the car and are not engaged in driving, you can start thinking of sending them content. You have sensors in the car, which track what you are talking and can talk back to you. It can send you suggestions as to where on your route can you have a good coffee or anything you want to eat, or can show you discount offers of items being sold in shops on that route etc. Various other such services can be provided by tech companies and derive an economic value out of it while collecting data about the passengers.
3. Artificial Intelligence (AI) :
There is a consensus amongst most technological players of the world that Artificial Intelligence AI is going to have a very major impact on the economy in near future that is just 5 to 10 years from now and there is a need to invest a lot in developing Artificial Intelligence, but the question is what is the business model for artificial intelligence ? Robotics is one thing that comes to mind but its market is not big enough to justify such a big investment. Chatbots may be another market, again it’s not that big to justify those big investments (which runs into billions of dollars to develop Artificial Intelligence). But when we think about Cars, It is something every human being would like to use it. It is an ideal platform for Artificial Intelligence, because if you want to have a car which is self driving you need sensors that understand the world at perception levels very close to that of humans, understands driving negotiations as to how we merge into traffic which should become a competitor to human intelligence. So now we are figuring out serious artificial intelligence use cases along with a business model. Since we now have a business model which deals with billions of dollars annually worldwide, it makes total sense to develop Artificial Intelligence (AI).
That is why Self Driving Car is becoming such a huge industry. It makes mobility on demand very attractive. Any organization who is into software development be it startup or companies like Uber, Google, Tesla, Apple, or any company automobile industry is working towards positioning itself as a player in self driving cars segment.
Artificial Intelligence in Self Driving Cars:
Lets us try to understand, what is the kind of Artificial Intelligence that we need when it comes to self driving cars and why it is exciting from a technological point of view ? In order to enable a car to drive on its own, we need to cover three technological pillars:
1. Sensing:
Just like we humans have senses such as eye, ears, touch, which collects signal and then these sensed signals are send to brain which then directs the body to take an action and so on. Same way we need to enable self driving cars to have sensors covering 360 degrees using cameras, sensors for redundancy like radar scanners and laser scanners then the data from all these sensors would go to a very powerful high performance computing device and the purpose for all this computing is to build an environmental model that :
a) tells where all the entities around the car are, such as pedestrians, cyclists, road turns, signals, symbols, barriers etc,
b) tells what these entities are they doing or what is their status,
c) tell about the path, such as if this path is a highway.
Sensing is the most complicated area but it is the easiest amongst the three.
2. Mapping :
Mapping should not be confused with Navigational map (Google map) instead, they are maps which are high definition which are very detailed. Once you position your vehicle inside this map at a very high accuracy (in the range of 10 cm), you know everything about the roadway, about the drivable paths, about the delimiter, etc, the only thing that you don’t know about is where the road users are. That data is given by Sensing. So how do you create these maps in terms of making it efficient, in terms of cost of creating these maps and how do you update these maps, because these maps should reflect reality in a very short time, that means the moment environment changes you would like the map to be updated within minutes, today maps are not updated at such a fast rate. This further bring us to using crowdsourcing to build these maps. This is very hard to do at the same time very much necessary. Without these detailed maps, you cannot reliably propel a self driving car.
3. Driving Policy :
This is where serious intelligence needs to be embedded into the driving platform. There is a need of intelligence in sensing like how the pedestrians look like, what are they doing and the same for vehicles and so forth, but Driving Policy is all about negotiation. It is all about building a strategy as when we merge into traffic, we negotiate and this negotiation doesn’t comes naturally, we take driving lessons to do that. This is a sophisticated negotiation and since it is already difficult for humans, it is going to be even more difficult for computers, because we are talking about broad intelligence which involves understanding which vehicles to give way to, which vehicles to take the way from, what are the road users doing to merge into traffic. Also Driving Policy changes with location since the traffic differ from city to city. This has to be taken into consideration in Driving Policy being fed to the system.
Self Driving as Multi Agent System:
Self Driving can be thought of as a multi agent system. It’s actually a game because we are strategizing, we are making decisions and there are rewards in this decision, there are short term rewards, we are optimizing long term rewards, just like we are playing chess, we can sacrifice a point because few steps later we want to take the queen.
Evolution of Self Driving Cars Industry:
The period from 2015 to 2017, Driving Assist would be used. Driving Assist is a technology used to prevent accidents. This technology can be used in order to navigate a vehicle hands-free but stay in the lane, but the driver needs to be alert. It is not really autonomous driving because the system can make a mistake. It doesn’t have 360 degree awareness. It doesn’t see everything. It is not designed to add all possible crash situations so the driver needs to be alert.
Starting from 2018, we are talking about Highly Autonomous Driving, we are talking about real hands-free driving but in limited scenarios only on major highways. Highly autonomous driving means the driver does not have to have eye on the road because the system will give a grace period for a number of seconds somewhere between 10 seconds to 30 seconds to take over and if the driver does not take over the system knows how to safely move aside and stop. In those roads you can take off your eyes from the road and that would start in 2018 on the highways.
The real action starts in 2021. Its called Fully Autonomous Driving . In other context it’s also called level 4 autonomy where we are talking about cars driving in city doing mobility on demand, kind of Uber without a driver.
My Name is Ashish @ashish_fagna. I am a software developer . If you enjoyed this article, please recommend and share it! Thanks for your time.
You can also contact me on ashish [dot] fagna [at] gmail.com