Self-driving city bus masters tunnels, poor road markings

November 29, 2016 // By Christoph Hammerschmidt
Driving in urban environments is particularly challenging for the electronic brains of an autonomous vehicle: Tunnels confront the cameras with changing lighting conditions, crossing traffic keeps the algorithms under constant alert. If the vehicle is a bus with a total length of 12 meters (40 ft), maneuvering becomes even more difficult. A team of scientists from the FZI research institute in Karlsruhe (Germany) and carmaker Daimler AG successfully sent a self-driving bus across such an urban parcours.

Without any intervention of the safety driver, the “Future Bus” mastered a test route of 20 km that included tunnels, crossroads with traffic lights, particularly dreary stretches (actually, the route led along a dyke), stretches with markings that were difficult to recognize – and a top speed of 70 kph. The test drives were successfully performed in July on Europe’s longest bus rapid transit line in the Netherlands between the Amsterdam airport and the town of Haarlem. Over three days, the bus (based on Daimler’s Citaro model) performed shuttle rides; the safety driver never had to join in. In tunnels, the sensors had to master not only extremely strong variations in the lighting conditions but also fading and, in part, non-existing GPS signals. The research team countered these challenges with novel sensor concepts and robust localization. Robust localization means that the vehicle knows at any given time exactly where it is – a basic condition to automated driving. Where human drivers can fall back on their experience and knowledge, a self-driving vehicle has only its sensors and algorithms. A bus with a length of 12 meters with its complex vibration behavior along three axes is more difficult to drive than a passenger car, even for electronic control units.

To make sure the bus is working properly, the scientists relied on a complex system of sensors which included lane tracing cameras and sensor fusion. Combining data from cameras, sensors and other signal transducers in a smart way yielded a very exact localization, enabling the bus to navigate accurately to within a few centimeters. Before the live test at the bus, the researchers tested the system extensively in computer simulations.

The electronic gear in the bus was based on Daimler’s autonomous truck, demonstrated some two years ago, but in several decisive aspects it has been further developed and optimized for buses. In addition, numerous features and functions have been added. “Public transportation