Thanks to digitalisation, artificial intelligence takes the wheel
Olli – the first autonomous, self-driving shuttle bus – is partly 3D printed, has a top speed of 20 km/h, and carries up to eight passengers. At the moment, Olli drives fixed routes on university campuses – and is sure to soon be seen in many a town centre. Is this a four-wheeled revolution or just hype?
Will Olli revolutionise scheduled public transport?
Wolfgang Bern: In Germany, Olli has now been in service for six months on the EUREF campus in Berlin. Five hectares in the heart of Berlin’s Schöneberg locality are seeing the genesis of a future-focused development centring on energy-optimised buildings, a local Micro Smart Grid and keeping operating costs low by using renewable energy sources. As the site houses internationally renowned corporations as well as the Technical University of Berlin, I would imagine there are between 1,500 and 2,000 people on the roads there every day. We are using the experience garnered on campus to make Olli ready for road traffic.
What makes an artificial intelligence a good bus driver? Where are its limits?
Bern: Our autonomous system has superior reaction times to humans, but there are still problems with object recognition. For example, I am not aware of any self-driving vehicle that can recognise an open manhole – they would all just drive straight in. Unfortunately, there are still many examples like this.
What about Olli’s social skills?
Bern: Our partner in this project is the IT multinational IBM. Olli communicates via an interface with IBM’s supercomputer, Watson, which analyses the data from Olli’s sensors. The autonomous system can answer simple technical questions – for example about the charge remaining on the battery. One can also call up recommendations for nearby restaurants or local sights.
When do humans have to intervene?
Bern: Essentially, humans always have to intervene when the software cannot – yet – depict a particular situation. Here is an everyday example: If a leaf falls off a tree, Olli perceives this as an obstacle. But Olli cannot differentiate between a leaf, a brick or a baby crawling across the street. So we cannot yet dispense with the safety driver. A teleoperator in the control centre can intervene remotely at any time and stop the bus.
Whenever automated vehicles are involved in road accidents, the question of liability arises.
Stefan Schulz: In Europe, the person who commissioned the vehicle is generally liable – i.e. the registered keeper. Even the onward march of automation will not change this initially. In the event of a faulty driver assistance system, the manufacturer’s liability also kicks in automatically. That is not new, either. But we need to assume that with the spread of driver assistance systems, manufacturers the world over will be subject to more liability claims after accidents than in the past. Olli is not yet approved by the German Federal Ministry of Transport. Until this approval is granted, Olli’s top speed is limited to 6 km/h. ERGO and Local Motors have thus only taken out public liability insurance for the test operations in Berlin. As soon as Olli is street-legal, motor third-party insurance will be negotiated.
Olli is meant for use in various countries – the USA, Germany, Nordic nations. Do the respective legal situations demand special insurance programmes tailored to the needs of each country?
Schulz: Absolutely. Liability regulations vary from market to market. Minimum cover limits in the USA are rather low compared with those in Europe, so product liability is often discussed. In Europe, the cover limits may be higher, but in the United Kingdom, for example, strict liability is not yet applicable for motor vehicles as it is in countries like Germany. A solution to this issue is currently in progress in the UK, however. Over the next few years, lawmakers and courts will need to deal with many unresolved issues and prepare commensurate solutions – both nationally and internationally. In parallel, the insurance industry must also develop the required know-how to be able to respond to new developments appropriately. Beyond this, legislators and the insurance sector will need to look into a potential rise in terrorism risks. This could certainly happen – for example, if there was no one to notice unattended items on public transport any more.
Bern: My experience is also that requirements vary in the markets. It is still very difficult to accommodate all the necessary parameters.
Let us look to the future: What percentage of the vehicles in our towns and cities will be self-driving in 2030? And how will this change the safety situation?
Schulz: I expect that in 2030 approximately 50% of all vehicles will be at least partially autonomous – meaning they will drive with the aid of adaptive cruise control or lane-keeping assistants. Outside special restricted areas such as airports or large factory compounds, wholly autonomous vehicles are certain to take another 20 years. Road safety will not necessarily increase as a result. During the project there have already been many cases of unreasonable road users jumping right into Olli’s path just to see if it really would stop in time.
Bern: I think it will be at least a decade before we see autonomous vehicles on the roads. This is due less to the technology than to the population’s acceptance and ability to learn. Just one brief example of a problem: How can an autonomous vehicle communicate with a pedestrian who wants to use a zebra crossing? Initial suggestions and model solutions for this have been proposed. But before all road users can learn new rules, the industry will need to agree on a standard. And that will take some time.