Everything you need to know about self-driving cars
How to look smart in conversations about self-driving cars
There are tons of news regarding autonomous vehicles, but what really happens in the area? How self-driving cars actually drive? What companies manufacture them? Why they’re still outside of mass market? Let’s find out.
What is self-driving car?
It’s a car with automatic control system, capable of moving from A to B without driver intervention.
How self-driving car operate
To reach a destination, a driverless car needs to know the route, understand its surrounding, follow traffic rules, and interact correctly with vehicles/pedestrians. To accomplish all of this, it relies on the following key technologies:
- Camera: to see things like road lanes and traffic lights
- Radar: to sense objects, especially the big ones, and determine their distances with moderate precision
- Lidar: also to sense objects as Radar, but with much better precision and 360-degree coverage
- AI (Artificial Intelligence): the brain of the car. Process data from all of the sensors, manipulate car and make decisions.
Autonomy levels
The company called SAE International made a step for autonomy levels standardisation, which all of the market leaders take into consideration:
- Level 0 — No Automation: Human controls everything — steering, braking and accelerating.
- Level 1 — Driver Assistance: The car provides some help with either steering OR accelerating. Cars with cruise controls fall into this category.
- Level 2 — Partial Automation: The car can control both steering and accelerating at the same time, with human drivers monitoring the environment and keeping their hands on the wheel. Tesla’s autopilot is an example of Level 2 automation (or any other car with adaptive cruise and lane keep assistance).
- Level 3 — Conditional Automation: The car will control the car and monitor the environment, but may request human driver to intervene in certain situations. Probably 2018 Audi A8 is able to achieve level 3.
- Level 4 — High Automation: Compared to level 3, the autonomous car can handle more complex driving tasks. It may still ask for human intervene in rare situations such as extreme weather. Otherwise it will gradually slow down and stop. OSCAR project is one of them.
- Level 5 — Full Automation: A truly self-sufficient car. It will not ask for human intervene and can carry out all driving tasks. Steering wheel is optional. Won’t happen earlier that 2023.
Key players in the market
Most of car manufacturers realized that self-driving cars are the future and rushed to buy self-driving startups and create new divisions. Besides of car manufacturers a lot of startups and IT-giants like Google or Yandex are also participate in this race.
General Motors
As one of the leading car manufacturers, GM has spent a lot of money to maintain its lead position in driverless cars. In 2016, it acquired self-driving tech company Cruise Automation for over $1billion. Moreover, trying to dominate the market, GM bought its own LIDAR manufacturer so that it can rely less on partnerships. GM has been testing in San Francisco with a plan to expand to New York City. It plans to commercialize its self-driving car with ride-hailing services in 2019.
Waymo
Being the oldest self-driving company Waymo was established a decade ago. Valued at $175 billion, Waymo has accumulated over 8 million self-driving miles with a fleet of Chryslers, Hondas, and Jaguars. Pledged to commercialize by the end of 2018, Waymo is leading the self-driving car race.
Uber
From the high-profile lawsuit against Waymo to the operation halt in Arizona and Pittsburgh, Uber has had a bumpy ride. However, Uber has not given up. With partners like Volvo and Daimler, Uber scored a $500 million investment from Toyota this year. Back on the streets of Pittsburg, Uber’s autonomous vehicle is driven manually to recalibrate its HD map of the city. It is also pumping money to its driverless car engineering hub in Toronto to continue its mission.
Lyft
Compared to Uber’s aggressive development and expansion strategy, Lyft’s approach is more controlled. Lyft partnered with Aptiv, a formerly bankrupted company that turned its tide by investing in driverless technology. Together they have completed over 5000 paid rides with 20 vehicles in Las Vegas, servicing popular destinations on the strip. When requesting Lyft rides, users can opt-in for driverless cars, and for now, there’s still a safety driver in the vehicle to monitor the ride.
Tesla
Tesla takes a completely different approach to driverless technologies. Instead of using LIDAR, Elon Musk believes the advancement of camera and image recognition technology are sufficient for automation. Although its current vehicles are equipped with auto-driving features, they still require human attention, and we have seen accidents reported due to the lack of human interventions. It’s yet to be seen how far Tesla can push its self-driving technology.
Baidu
Baidu leads the biggest driverless car effort in China since 2014. In 2017, it announced the Apollo, an open source project to facilitate the research and development of driverless cars. Baidu aimed to mass produce autonomous vehicles by 2019 to 2020, but their ability to successfully carry out this plan becomes doubtful when it lost multiple key AI talents and COO Lu Qi.
What’s taking so long?
Waymo was first established in 2009, and almost a decade later, its cars are finally getting ready to become commercialized. That’s quite a long time to release a product. Why does it take so long? Although eager to advance as quickly as possible, the self-driving car industry is facing key some challenges that they need to overcome in order to truly become common on streets. These issues include:
Lidar
LiDAR fires rapid pulses of laser light at a surface, some at up to 150,000 pulses per second. A sensor on the instrument measures the amount of time it takes for each pulse to bounce back. Light moves at a constant and known speed so the LiDAR instrument can calculate the distance between itself and the target with high accuracy.
However, even though Lidar is not a new technology, it still needs further development. The bulky equipment comes with a hefty price tag of $75k per piece and is not ready for mass production. Now, over 50 companies are trying to improve Lidar and make it cheaper for mass production.
AI
We all have high expectations for what AI can accomplish, but it’s still far from perfect. For example, in the fatal accident caused by Uber’s self-driving car, the algorithm failed to recognize the victim, causing a delay in reactions. Furthermore, there are so many moving parts (literally) on the street, producing millions of possible scenarios. A tremendous amount of data is needed for the system to learn how to handle these scenarios, and we are just simply not there yet.
Weather conditions
Let’s put cards on the table, almost no autonomous car can operate in blizzard or heavy rain. But there is one exception — MIT. They managed to use localizing ground-penetrating radar (LGPR) for car localization.
HD Maps
Regular digital maps are insufficient for self-driving cars, besides that GPS accuracy is too low (3–10 meters). Meanwhile, autonomous car needs to localise itself with centimetre precision. Despite of bunch of sensors, car still required to have precise information about environment (road boundaries, lanes geometry, traffic signs, etc.). All this data is presented in so-called HD maps.
In order to keep the data up-to-date companies use mapping vehicles (special cars with 360-degree camera and lidars) to collect road data from time to time. So with self-driving race also mapping race comes forward with companies like Here, TomTom, DeepMap, lvl5, Carmera and others.
Infrastructure
At the moment AI using machine learning techniques (deep learning) have been the focus of development for autonomous vehicle software. A communications protocol layer for self-driving cars can add further improvements for SAE Level 5 compliance. It is a protocol that supports V2V (Vehicle-to-Vehicle), V2I (Vehicle-to-Infrastructure), V2P (Vehicle-to-Pedestrian), V2H (Vehicle-to-Home), V2N (Vehicle-to-Network) and V2C (Vehicle-to-Cloud) communications. This will allow self-driving cars to communicate and share information among themselves. This will make self-driving cars aware of each other much like how people interact.
V2V Vehicle-to-Vehicle — Allows V2X enabled self-driving cars to communicate with one another.
V2I Vehicle-to-Infrastructure — Allows self-driving cars to get information from buildings, bridges, roads, traffic lights etc.
V2P Vehicle-to-Pedestrian — Makes use of pedestrian detection systems that can work with a car’s ADAS.
V2H Vehicle-to-Home — Smart homes can send and receive information directly from the car.
V2N Vehicle-to-Network — This is a mobile connection from the car to a carrier’s cellular network.
V2C Vehicle-to-Cloud — Provides direct access to cloud networks using secure TCP/IP connections.
Human trust
While the industry has slowly got a grasp of human-vehicle communication, human trust is a more complex problem to solve, and it directly affects the adoption rate. In a research conducted by the Reuters and Ipsos, only 38% men and 17% women surveyed said they would feel comfortable riding a driverless car. Another Research shows that to gain people’s trust in autonomous vehicles, the industry needs to demonstrate the cars’ reliability, performance, privacy (e.g., location tracking), and security (e.g., hacking).
What’s next?
We’re wintessing how self-driving cars slowly appears on our streets. It’s highly unlikely that it will become a mass thing - neither algorithms, nor infrastructure are ready. Although, with V2V/V2I arrival we could see special zones for autonomous vehicles, where you’d be able to summon a robo-taxi.
At StarLine we’re building autonomous future of Russia. If you want to work on amazing products in a motivating environment where you can grow and lead, StarLine is the right place for you.
Stay tuned, a lot of exciting things are coming.
Twitter: https://twitter.com/starline_oscar
Website: https://smartcar.starline.ru
