The CES 2017 fair in Las Vegas saw the BMW Group announce its intention to release a succession of prototype vehicles during 2017 in collaboration with Intel and Mobileye. These will form a fleet of 40 highly automated and fully automated vehicles by the end of the year. Test drives will take place on public roads and focus on two main types of use: driving without oncoming traffic (motorways) and driving in city centre environments. The test drives will be mainly conducted in the home countries of the three partners, namely USA, Israel and Germany.
By developing these BMW 7 Series advanced prototypes collaboratively, the partners will ensure the timely roll out of the BMW Group's first highlyautomated series vehicle (level 3) – the BMW iNext, due in 2021. BMW iNext is the BMW Group's first venture into highly-automated driving. From a technical perspective, the BMW iNext will also be capable of level 4 and 5 operation. Whether or not this is achievable in practice depends on a number of external factors, but it is not yet possible to predict how these will develop.
For an autonomous vehicle to be considered market ready, it must behave safely and reliably in any conceivable driving situation, as well as operating in a way that is predictable for other road users. Theoretical calculations have determined that around 240 million kilometres (150 million miles) of testing on public roads would be needed to provide assurance for every situation. In practice, this is neither practicable nor sensible. In fact, the most relevant tests relate to a much smaller number of critical driving situations, not the total distance travelled. Instead, autonomous vehicle safeguarding is carried out by analysing "foundation" situations that have been investigated in real-world trials. These situations are then extrapolated using stochastic simulation to provide comprehensive validation. For example, in future BMW will be in a position where it is able to test around five million driving situations per simulation for every software release within a very short space of time.
Artificial intelligence is a discipline within the field of computer science. Its goal is to use computer programs to solve problems that could not otherwise be solved without using the intelligence of a human being. Artificial intelligence is important as a key technology for many aspects of mobility, now and in the future.
There are many different areas at BMW where it is being applied. These include optimisation of production processes and the development of customised natural-language interactions for customers. Another field where artificial intelligence can be applied is in the creation of highly accurate road maps with dynamic content, such as temporary obstacles and live traffic information. It can also play a key role in intelligent multimodal routing, intelligent car sharing and ride sharing, provision of location-based services and other services that are personalised based on user context.
The BMW Group is already active in all these areas and is working on combining them into an total user experience that is both attractive and useful to users. Artificial intelligence is increasingly allowing computers to find solutions to highly complex problems, something that would have been inconceivable just a few years ago.
Software developers at the BMW Group are playing a significant role in such developments and have the opportunity to experience the new technology directly through the product.
Artificial intelligence as a key enabler for autonomous driving. It was clear even in the early days that autonomous driving would not become a reality if purely rules-based approaches were used. Instead, realising the vision of autonomous driving requires machine learning systems.
Communications A diverse range of real-world data must be collected by a vehicle's on-board sensors in order to facilitate a data-driven development cycle. This results in vast quantities of data that must then be processed and made available by the artificial intelligence system. A data centre is currently being set up for this purpose, in collaboration with Intel, and will be further extended in the coming months. Training of neural networks and further development of algorithms requires the data to be always quickly accessible, so the facility is being equipped with a corresponding amount of computing power. The data centre will also simulate scenarios that occur so rarely in the real world that test coverage could not otherwise be truly comprehensive.
The result is artificial intelligence with an ever-increasing ability to develop models of reality. Another artificial intelligence system is required in the vehicle to make an intelligent interpretation of the situations it faces based on the models. Without this, the vehicle cannot derive a driving strategy with the necessary degree of confidence.