Robots will make fatal mistakes, Continental concedes

People killed by autonomous vehicles are an inevitable price to pay for a Vision Zero future.

That’s the admission made by Dr Elmar Degenhart, chief executive officer and executive board member of Continental speaking to TU-Automotive at 2018 NAIAS.

When asked to reflect on comments made to us by Lockheed Martin’s Michael DeKort in the article Guinea-pig testing casts a ‘shadow’ over driverless tech, Degenhart said the greater goal justified the mistakes that any new technology will make.

He said:“No technology is perfect. We might see cars that are driving at Level 4, Level 5 which will not be able to avoid an accident. We need rules for our engineers who are programming the code for many conflicting situations, yes definitely.

“We have a committee in Germany, for example, which is taking into account these ethical questions. So we have to take advice not to leave it up to the engineers alone how to program and how to take care on such potentially conflicting situations.

“Should we not go for these technologies because we are at risk that automatic driving vehicles might kill people because we are afraid of that? And not taking the chance to help hundreds of thousands of others to save lives? This is not an option from our point of view.”

Degenhart believes the advancement of autonomous technology will, ultimately, create a much safer world with fewer casualties and greater ecological efficiency. He explained: “First of all our strategy of Vision Zero that we came up with is that we are going for zero fatalities in the first step and zero accidents in the second step.

“Now, we’ve done a rough calculation and we estimate that if you were to equip all vehicles around the world with the technology standard that exists in the US, for example such as safety belts, having stability management control, having a multitude of airbags inside the vehicle, having emergency braking systems we believe you could avoid 1Bn accidents doing nothing more than industrialising what we have already.

“Another 300,000 accidents could be avoided with sophisticated systems in the vehicle that can detect the environment much better than the driver can do. So we are working on this technology together with our customers and we will bring automated driving functions into the field Level 2 today going out Level 3 at about 2020 plus or minus a little, then Level 4 about 2025 and full autonomous probably between 2025 to 2030.”

His press presentation at the show highlighted Continental’s current work in electrication and autonomous vehicle solutions in the shape of its CUbE (Continental Urban moBility Experience) project with the French company EasyMile and its concept BEE (Balance Economy and Ecology) swarming pop concept. Speaking of CUbE he said: “This vehicle drives up to 40kmh or 25mph maximum speed. We have a close relationship with EasyMile and we could equip not only prototypes like this but also production series vehicles with electrified drivetrains, with connectivity technology to allow the vehicle to communicate with infrastructure, with the Cloud and with other vehicles within its environment and also with a complete system that will let the vehicle drive autonomously because these vehicles no longer have any steering wheels inside.

“They are good to transport up to 10-12 passengers and we have prototype applications running already in different cities. It’s extremely helpful with low speed applications which will be the first application of autonomous vehicles inside zones where only pedestrians are allowed for example. In this way the complexity of driving is somewhat limited compared to free driving on a highway.”

Degenhart sees the BEE concept taking the driverless technology a step further as transport for a smaller number of passengers. “We will be able to use these vehicles in a form of a swarm concept where we could allow, two, four or a set of six to eight vehicles to run very close and in parallel. We would link them together [wirelessly] and we would use them as a form of public transport for the future.

“The distance we are scheduling and going for would be in the range of 200 miles per day and the speed we could imagine to allow for 40-45mph. If you take the latest projections that by 2050 two-thirds of the world’s population will live in big cities, then we will certainly see thousands of these kinds of vehicles in our cities.

“They would be very flexible and agile and could be used by handicapped people very easily but what is necessary is for the infrastructure of cities would have to be designed in a way that would make driving for these vehicles easier.”

However, on the near future issue of driverless technology safety and regulation, Degenhart said there are still many questions that remain to be addressed. He explained: “Another part of the question is how to validate the technology? It will be impossible to take care of each and every situation you can think about outside of the vehicle driving around in, say, Level 4 automatically or even fully autonomously. Here artificial intelligence algorithms are coming into play. You have to allow the car to achieve self learning experiences because you cannot foresee every situation you could possibly think about.

“The question is how to validate self-learning algorithms? We have no answer for this today. Therefore, we say if you believe you can implement Level 4 driving the in the 2020 timeframe, this is completely unrealistic. The authorities have the responsibility [to ensure] that the automotive industry is proving it has control over validation processes [because] they will not allow things that are [im]purely validated.

“We have to find solutions for these questions and we don’t have all the answers yet.”


Leave a comment

Your email address will not be published. Required fields are marked *