German Professors Analyze UBER Accident

29/03/2018

Contact

Name

Lutz Eckstein

Institute Director

Telephone

work Phone
+49 241 80 25600

E-Mail

Professors from RWTH Aachen University, the Technical University of Munich, Ulm University, Braunschweig University of Technology, the Karlsruhe Institute of Technology and TU Darmstadt weigh in on the autonomous vehicle accident in the United States.

One of UBER's self-driving test vehicles strikes and kills a pedestrian during a nighttime drive – neither the autonomous vehicle nor the test driver reacted. Even though there was an emergency backup driver behind the wheel, this tragic incident has resulted in the first pedestrian fatality associated with self-driving technology to date. Naturally, many questions are being raised in public discussion concerning the purpose of and accountability issues with automated driving technology. Professors from German Universities who have been researching autonomous driving for many years already and are working together in the Uni-DAS e.V. association, have likewise analyzed the accident.

  • Professor Klaus Bengler from the Technical University of Munich
  • Professor Klaus Dietmayer from Ulm University
  • Professor Lutz Eckstein from RWTH Aachen University
  • Professor Markus Maurer from the Braunschweig University of Technology
  • Professor Christoph Stiller from the Karlsruhe Institute of Technology, KIT
  • Professor Hermann Winner from TU Darmstadt

A few days after the spectacular fatal accident the UBER-vehicle was involved in, a video of the accident taken by outer and inner cameras was publicized. It clearly shows that the vehicle did not slow down, nor did it initiate any kind of evasive maneuver, just as the police had already reported. It instead drove into an area with a posted speed limit of 30 miles per hour going 40 miles per hour and it struck the pedestrian at full speed, since not even the safety driver reacted.

The operation of the autonomous vehicle by UBER in the present case can be characterized as a form of category 2 automatization, even though higher levels of automatization had also been tested. This means that the safety driver sitting in the vehicle was supposed to oversee all functions at all times and when in doubt take over control. Why she did not do this and why she constantly averted her gaze from the road, cannot be ascertained from this video alone. Due to the somewhat poor quality of the video, it furthermore cannot be said what the safety driver should have seen had she been attentive. Therefore, it remains to be clarified over the course of the investigation, why the safety driver neither reacted in any way nor corrected the excessive speed of the vehicle. The detailed analysis of the accident by authorities must furthermore show whether safety drivers are being sufficiently trained by UBER and whether the safety concept in place for test operation was adequate.

Accident hard to comprehend

The circumstances of the accident are particularly hard to comprehend in light of the current level of technology. The automated vehicle owned by UBER was equipped with lidar, radar and camera sensors. Even though camera sensors may have limited perception capability in the dark, it needs to be said that the pedestrian was already clearly visible in the camera's image for more than a second before the collision. Lidar and radar sensors are even considered active sensors , which means that they actively send laser pulses in the infrared range, or radar waves, and thereby – on the basis of reflections – measure distances to objects, their relative speed, and their size. These sensor principles then are obviously guaranteed to function in just the same way in the dark. An emergency braking system, which has already been installed in many production vehicles to date, functioning with the help of these sensors would have at least warned the driver and slowed down the vehicle – and as a result reduced the extent of the collision or possibly even have prevented it.

A setback for automated driving

The accident in itself is currently regarded as a setback for automated driving. Deadly collisions with vulnerable road users, such as pedestrians or cyclists, are, however, not a rarity even in non-automated traffic in Germany. Approximately 50 percent of road traffic fatalities are in the particular group of road users that are hard-to-see in the dark. Furthermore it has to be noted that no technology – neither existing nor future systems – can offer 100 percent complete security. Unfortunately this is also true with regard to mobility – a fact that is documented by the number of injured and killed in road traffic, rail traffic, and air traffic.

Even though simulation plays an essential role in the development and safeguarding of automated driving functions, automated driving tests on public roads will continue to be necessary, because the complexity of real traffic situations can never be modeled in simulations or on test sites entirely.

Social acceptance of the technology hopefully not diminished

In the medium to long term, it is expected that carefully developed and secure automated vehicles will lead to a decrease in the number of accidents compared to today. That is why the authors continue to hope that this unfortunate accident will not result in a decrease of social acceptance for this technology. In order to pave the way for these future benefits to society, however, it is important to realistically evaluate current capabilities, technical and technological constraints, as well as tough safeguarding processes.

Start-ups in the USA in particular appear to be under enormous pressure to succeed – also due to the fact that they have to compete for financial investors. This may then lead to premature testing on public roads and premature showcasing of their product. As long as companies strive to outdo each other in announcing who will be in the best position to open up the billion-dollar market with driverless robot taxis, then this pressure will also bring with it a higher risk of premature testing on public roads.