The roadmap to your self-driving family car – Last year saw definitive changes in the self-driving vehicle landscape. In utopian future models, autonomous vehicles are regarded as the base of a sharing economy. 2017 proved that such a system is a long way down the road. The reason behind this is extremely simple. Autonomy is expensive. The cars, the sensors, the development time, the know-how and expertise all add-up.
This, in turn, has a rather obvious result. Current self-driving cars are not systems that everyday families could afford. The general public will first meet autonomy in two different models. The first is (local) government subsidized public transport. Similar systems are already being tested in Las Vegas, Singapore and other cities around the world. The second model comes from asset-heavy players within the mobility industry. It aims to control the taxi ranks of the largest metropolitan areas around the world.
However, autonomous technology is making its way to the average consumer. Nevertheless, the technology, and the price to be paid will determine common use. Current lidar-based sensor setups are prohibitively expensive. This will change in the future, as the technology advances. However, as lidar point cloud resolution increases, processing the data will become very similar to processing camera-images. As a result, all approaches to autonomous technology face the same challenges in the long run.
General consumers will first experience limited autonomy in the form of ADAS functionalities. These will appear in high-end models from top OEMs, with lower tier models following hot on their heels. To increase road-safety lawmakers may even make elements of self-driving technology compulsory for all road cars produced after a certain date. These could range from already common collision avoidance systems to full highway autonomy.
Safety is the most important factor to be considered. Over the course of 2017 the autonomous industry accepted that simulation technology will be vital to achieving safe autonomy. The millions of road miles covered by prototypes may be false indicators. A self-driving artificial intelligence which has covered each road in California several times over would be nearly useless in a snowy climate.
However, diversifying road-testing is logistically demanding. Low-key estimates state that a self-driving system would have to cover 5,000,000,000 miles (8,000,000,000 km) in testing to be safe. That is at a minimum 150000 car years (a fleet of 150 self-driving prototypes would complete the tests in 1000 years). Calculating with two engineers in the car that means at least $18 billion. Through concentrating on more interesting scenarios, and utilizing the scalability of computer systems simulation technology can reduce this number to 10000 simulator years, and a cost of about $125 million. There is still road testing to be done after simulation testing, but the numbers are nowhere near so extreme.
On a general level, in simulators, systems can cover millions of miles in one night. They do so in different locations, changing environmental conditions, in varying amounts of traffic. Not only does this make autonomous cars safer, it accelerates their development. With a server park offering engineers constant feedback on their work flaws and weaknesses are found quickly. Ideas can be tested straight away, and positive outcomes brought to fruition immediately.
While autonomy still faces a myriad of questions, the public will first interact with it within this framework. The shift to autonomous mobility will be far from sudden. At a point, full-autonomy will happen, in an everyday production car, at a price-point that is affordable to average families. That's when we can start building the dream of global autonomous mobility.