Look at your car. It might be the last one you drive on your own. All the leading car manufacturers promise that driverless cars will conquer the world between 2018 and 2030. One small step aside the wheel for a human becomes a giant leap for the automotive industry.
Roadmap to autonomous driving
Self-driving car technology liberates you when you’re on the road. Intelligent mechanics can save fuel, reduce emissions, ease the flow of traffic, improve transportation and supply services. Moreover, autonomous cars promise to save millions of lives by minimizing accidents caused by human error.
Functional architecture of autonomous cars
Roadblocks to the driverless future
The biggest challenges to autonomous vehicles reaching the global market fall into three categories: legal, ethical, and technical. Legal and ethical challenges are hard because their solutions are outside the direct influence of the automotive industry. At the same time, technology is improving faster and faster as more companies join the race to create autonomous cars.
Technical pillars to autonomous driving
Today, we’ll take a closer look at the five technical pillars that stand tall for self-driving cars to come in the immediate future.
Car sensors and software
Self-driving car navigation is based on four key elements:
A few years ago, sensor quality and cost were the hot spot of every discussion about the driverless future. Google used a LIDAR system priced at around $75,000 per unit, which was half the total cost – around $150,000 – of each of their driverless vehicles. These prices made self-driving cars a dead weight for the global market.
Since then Velodyne, a company specializing in LIDAR production, has started offering a small but capable unit for just $500. So, the issue of cost as the main roadblock to autonomous vehicles has gradually started to fade away.
At this stage, the responsibility for our self-driving future rests primarily with software providers. They have the entire niche to develop an automotive expertise that would meet the following autonomous car requirements:
Navigation and mapping
Relying on sensors, software, and GPS data, an autonomous vehicle processes nearly 4,000 gigabytes of data daily. Advanced machine learning algorithms need timely updated data to react to the ever-changing environment. A continuous exchange of data between vehicles and the cloud can overload data networks and data storage, creating additional challenges for software vendors.
Currently, map providers are putting every effort into reducing the volume of data transferred between a vehicle and the cloud by trying to maximize the amount of data processed within the vehicle itself. One proposed solution is a so-called “self-healing map” that would use on-board AI algorithms to navigate without stepping too far to the cloud.
Built-in computer vision algorithms could recognize landmarks and road signs picked up by cameras. This would allow a vehicle to distinguish static objects in the changing environment and generate a detailed 3D picture within the embedded navigation system. Reducing the amount of data exchanged with the cloud would save the data connection primarily for exchanging missing map information among drivers. This would encourage map makers and OEMs to prepare for a new era of crowd-sourced maps.
Machine learning and open platform
Most innovative tech companies have introduced their own open location platforms (OLPs) that let a car literally teach itself how to navigate. An OLP can help a car manufacturer equip their current models with advanced driver assistance technologies right away instead of having to design new prototypes. This cross-industry cooperation between tech companies and auto makers might dramatically speed up the appearance of driverless cars.
Collaboration between OLP providers and car makers gives companies the opportunity to collect as much traffic data as possible. With a growing number of cars equipped with OLP, sensors, and powerful AI solutions, the volume of data on the road environment will increase exponentially. Instead of driving endless kilometers worldwide with hundreds of mapping cars to collect traffic data to produce high-resolution maps, millions of regular cars connected to an OLP will provide a detailed picture of the world in record time.
Moreover, with access to traffic data from various sources and industries, an OLP has a much greater potential. An open location platform can connect to a city’s infrastructure, merchants, and suppliers to become one large data hub for map users and drivers. By analyzing where a driver has previously driven, product companies can suggest deals for a potential customer and not just the right turn to take to escape a traffic jam.
The accelerated improvement of HD maps will take autonomous driving to the home stretch. Raw data from sensors can be accumulated as a common source of easily accessible data about routes, traffic situations, and environmental changes. The key to landing all this big data in one place is the open location platform.
Connectivity and communication
Would you be ready to trust your life to an AI without even being aware of how it behaves? The issue of placing trust in a vehicle can be resolved with a car’s communication abilities. A vehicle should be connected not only to the cloud (to receive geolocation data and map updates), but also to humans (V2H), other vehicles (V2V), and infrastructure (V2I) in order to be trustworthy traffic participants.
Vehicle-to-human (V2H) communication is based on a human-machine interface that enables the driver to observe the vehicle AI’s intention before it makes a move. Based on natural language processing algorithms and voice recognition, V2H communication can create an extra link between self-driving cars and their owners.
Vehicle-to-Vehicle (V2V) communication sets autonomous cars on the most efficient and safest course. By communicating with other vehicles, a self-driving car can receive information about upcoming obstacles, traffic congestion, and pedestrians on the road long before they’re in front of the vehicle or appear on the map.
Advanced V2V communication systems will make it much easier for autonomous vehicles to understand traffic behavior. Essentially, this is because vehicles will communicate in their own language among each other. Currently, a vehicle’s AI has to adapt to the human-made system of signs and signals to predict the movements of other human-driven vehicles.
Vehicle-to-Infrastructure (V2I) communication allows an autonomous car to exchange data with its environment, including with buildings, traffic lights, and road signs. For example, when driving too fast, a car might receive a signal from a traffic light about a possible red-light violation. Or a car might receive a signal about a potential green line for further 30 kilometers if it drives at 50 km/h.
The concept of V2I connectivity brings us closer to smart cities that can spread data-driven decision-making even further than the automotive industry.
Cybersecurity and autonomous cars
Driverless cars are already very smart. But are they smart enough not to be hacked? The answer is that they need to become that smart.
An advanced driver assistance system (ADAS), for example, can stop your car when you’re driving into a wall. But do you want your car to stop without any reason only because of a hack attack? The significance of cybersecurity rises dramatically in a world where a hacker can drive, stop, and control your car remotely.
Properly built security systems can be a strong barrier to halt a thief. If a car’s system detects an abnormality in the vehicle’s behavior, it can block data access or even cut the car connection. In addition, preliminary penetration tests and prediction of possible threat scenarios make the security burden much easier.
Tesla driverless car was intentionally attacked several times a day with the reward ranging from $100 to $10,000 only to find a critical bug or any other potential vulnerability. In one week, Tesla Model S was hacked in a variety of ways, allowing attackers to remotely open the door, change the music, control the air conditioning system, and even start the vehicle. There are many similar examples from other manufacturers. Finding these security holes out in the field conditions can help make autonomous cars more secure.
Car makers and software security providers are forced to work together to prevent risks. The best strategy is to create an advanced architecture capable of detecting and preventing penetration or any conflict with standard vehicle algorithms.
Despite the roadblocks, autonomous driving is quickly approaching the global traffic. It promises to change our perception of urban mobility, mapping, supply services, smart cities, and AI capabilities in a larger sense.
In this race to win the driverless future, the tech part plays the role of a catalyst for a wider span of industries to consider autonomous driving as the most competitive though promising field in the nearest years. Autonomous driving is making the automotive industry the benchmark for other industries that are relying on connectivity, embedded solutions, and deep AI integration to drive the future.
At this point, one small step aside the wheel has already been taken. The giant leap that will change the world has just begun.
Intellias is ready to become your trusted partner in autonomous driving. Get in touch to create the driverless future together.