Autonomous driving technologies are unanimously seen as the future of the mobility industry. Automotive experts are in full agreement that as soon as they become safe for widespread use, self-driving cars will turn the auto sector upside down. Traditional carmakers are joined by the largest tech companies and financial investors, driving innovation in the race to attain true (Level 5) autonomy as fast as possible.
But self-driving cars obviously aren’t a single piece of technology. Rather, automotive systems developers are working on a number of technologies that all contribute to vehicle autonomy.
Developing self-driving technology
To name just a few tech giants joining the competition to deliver autonomous driving technology, Google, Apple, Baidu, Intel, Microsoft, and Samsung are now all involved in self-driving development, along with carmakers like BMW, Volvo, GM, Daimler, Toyota, and lots of others. Despite all the financial power and expertise that these companies bring to the table, there are still several technological challenges before we can safely take a nap behind the wheel.
Better known technologies to enable self-driving cars include ADAS (Advanced Driver Assistant Systems), V2X (Vehicle-to-Everything) communications platforms and network connections, or Electric Vehicle & battery technologies. One way or another, all of these are seen as stepping stones to real vehicle autonomy.
Related reading: The Key Enabler Technologies of Autonomous Driving
But there are quite a few other technologies that are deeply embedded in autonomous development. Belonging "under the hood", they might not make the news as often as self-driving cars themselves, but these technologies are vital to attaining true vehicle autonomy. Taking a closer look at them helps understand the complexity of modern mobility solutions.
Unseen technologies enabling autonomous driving
Vision & image processing
Self-driving cars rely on a whole variety of sensors, cameras, radars and Lidars to gather information on their surroundings. To provide accurate position and orientation information, autonomous cars will need to combine all the data from these sources in an intelligent way, which is why sensor fusion is also a hot topic.
But let’s not forget that our driving environments have evolved to be a primarily visual business – without good eyesight, you’re likely to get in trouble on the road. Vision processing is an up-and-coming technology that will help provide autonomous cars with something resembling that inherently human ability.
Advanced software algorithms, simulation software
The vast amounts of data generated by all those sensors in a self-driving car will need to be managed and processed. The embedded computing infrastructure that serves as the autonomous car’s “brain” will have to analyze all that data, and make decisions in real-time. In order for the car to make accurate decisions on when to brake, accelerate, or change lanes to keep us safe, the software algorithms that contribute to that decision have to get very smart.
In the IT world today, “smart” essentially equates to “based on Artificial Intelligence”. And machine learning and AI really do contribute to the development of autonomous solutions. One way that self-driving algorithms are trained today is via simulation. The software governing the car’s actions is virtually placed into thousands or millions of traffic situations, where its actions and the results of those actions help fine-tune the operation of these autonomous decision-making units. Simulation is one of the most important tools that developers today use to ensure safety in tomorrow’s self-driving cars.
Related reading: The Road Ahead: 2019 Trends in Automotive Technology
Navigation and mapping
When we’re driving along a familiar route, most of us don’t need to turn on the GPS. For a car that’s driving autonomously, there is no familiar route, and orientation is vital: the software making its decisions needs to be tightly integrated with navigation and mapping technology. The main challenge here is that existing maps were developed for humans, and require extensive optimization before self-driving cars can really work with them.
Your future car won't be able to simply turn on Google Maps, so navigation companies today are working on information-rich, HD maps specifically geared for autonomous vehicles. These maps will need to contain all the static objects, road signs (and their meaning), and even lane markings at a centimeter resolution.
Internal Networking and Buses
With all that data and software come hardware needs. For different applications, self-driving cars will contain several interconnected networks, and data will have to be transmitted among these electronic systems within the car. With traditional internal buses, bandwidth is becoming a challenge, not to mention that internal networking solutions designed in the past lack some of the features (such as security measures) necessary today.
The advanced automotive networking solutions being developed today will have to enable the fast and secure exchange of data within the car with super-low latency and a robust architecture to function reliably in any condition.
User Interfaces (Displays and Controls)
Since real self-driving cars are imagined as a 100% hands-free business, it’s no surprise that most of the tech challenges here are related to computer-to-computer or “backend” environments. But tomorrow’s passengers (today simply known as drivers) are still at the centre of vehicle autonomy.
Despite the mounting complexity of electrical and software components, passengers will require that cars are easy to use, and that they provide a seamless experience throughout the journey. User interfaces in today’s cars are mostly considered inferior to those applied in other everyday electronics products like smartphones. Even with today’s advancements (like head-up displays, haptics, eye tracking and motion recognition technology), there’s much development to be done on the user-friendliness of cars in general, let alone autonomous cars packed with high technology solutions.