Autonomous cars are inevitable in the automotive industry. Most of the automakers claim that they are able to launch autonomous cars in 2020, Tesla, confident that the company can do it earlier, in 2018. So, what is exactly autonomous car and what makes it so special?
Autonomous cars are divided to five types:
Level 0: It is the type of cars which are bought before 2011 whereby drivers need to be in complete control at all times.
Level 1: It is the type of cars which some functions such as cruise-control, automatic braking and lane keeping can be done automatically, however, the functions are working by itself. Drivers are still in complete control, which means that his hands and feet still need to be on the steering wheel and pedals all the time. Level 1 cars can be found in most of the cars on the road nowadays.
Level 2: It is the type of cars which some functions have already able to work together. The drivers are allowed to keep their hands and feet off the steering wheel and pedals. However, drivers still required to be alert all the time. Tesla Autopilot is categorized in this level of self driving cars.
Level 3: It is the type of cars which most of the functions are enough for the cars to be able to drive itself automatically. Volvo Intellisafe Autopilot belongs to this level of self driving cars.
Level 4: It is the type of cars which all of the functions of the cars are sufficient for the cars to operate itself. Drivers are not expected to be alert all the time. On top of that, this design is suitable for both occupied and unoccupied vehicles. Tesla cars and Google cars aimed to achieve this level.
There are definitely challenges that automakers and technology companies need to overcome in order for the autonomous cars to be able to launch and used on public roads. One of the challenges is in the legal hurdles and regulation insurance. Last December, Department of Motor Vehicles in California released draft regulations that require drivers to sit behind the steering wheels and also require the vehicle to be equipped with steering wheels and pedals too. Google, on the other hand, plans to design self driving cars without steering wheel which not aligned with the authority.
Insurance is also becoming a challenge for self driving cars. For example, if the autonomous car is driving to a blocked road on a one way street and the only way to get out of it is by u-turning and unfortunately, it causes accidents. Therefore, who is responsible for it, the owner or the car manufacturer? Therefore, unless these challenges have been resolved, then autonomous cars are not ready to be taken on streets.
Another challenge is more on technical challenge as how to make autonomous cars to be able to handle all kinds of scenarios without human driver interference. One way is by equipping autonomous cars with advanced technology such as deep learning and advanced hardware such as sensors or cameras for them to be able to do things which they are designed to do.
Deep learning is an algorithm whereby the system is able to learn and handle new scenarios which never encountered before during testing phase. Therefore based on the data which is collected from the sensors and cameras, the system is able to find patterns, predicts the next thing which is going to happen and find the best way to handle the scenario. All is done without interference from human driver.
To develop a deep learning algorithm, major automakers are working together with another technology company, Mobileye. The company has developed smart cameras which are able to detect the surroundings, calculates the possibilities of a collision and stop it from happening. Mobileye has worked with Hyundai and Tesla among other automakers to develop vision related advanced driver assistant systems.
Deep learning algorithm requires an enormous processing power which normally known as supercomputers. These supercomputers are no longer taken form as big bulky machines. They are more likely build as a microchip. Nvidia is one of the companies which has experience in developing them.
Therefore supercomputers should be able to collect big data collected from cameras, sensors or lidar systems, analyze the data in real time and then make decisions based on it. All need to be done in shortest time possible. The input can be as big as multi terabytes in size. It should be able to handle the input taken when the car is travelling on a six-lane highway with thousands cars travelling surrounding it and at different speeds. It should be able to handle thousands of objects located on the roads.
Hopefully, both authorities and automakers will find way to resolve the issues so that we can experience the benefit on having self driving cars on our roads.