- Mid-term review in the UR:BAN research project
- Safely and stress-free through the city in the future
- Stixel world makes detailed modelling of the surroundings possible
- Lane Change Assist identifies and factors in driver intention
UR:BAN (German acronym for Urban Area: User-Focused Assistance Systems and Network Management) is the first research initiative to address complex urban traffic situations. At the mid-point of the initiative established in 2012, Daimler AG together with the 30 other partners involved in the initiative presented promising results in Brunswick on 14 May. The goal of the projected funded in part by the German Federal Ministry of Economics and Technology (Bundesministerium für Wirtschaft und Energie) is to develop intelligent and cooperative driver assistance and traffic management systems to make urban traffic safer, more stress-free and more economical. Prof Dr-Ing Ralf Guido Herrtwich, Head of Corporate Research and Advanced Engineering Driver Assistance Systems of Daimler AG, emphasises: “The UR:BAN initiative perfectly matches our “Real-Life Safety’ strategy, with which we want to prevent accidents and mitigate their consequences. The intelligent fusion of sensors and systems helps in detecting potential dangers in urban traffic and reacting to them in a timely manner“.
Early detection prevents accidents
The focus of Daimler’s involvement in UR:BAN is the ‘cognitive assistance’ research pillar. The objective: new assistance systems are supposed to provide driver support in complex situations such as at junctions with pedestrians and cyclists, in tight spots, with oncoming traffic and when changing lanes, for example. Herrtwich adds: “With our involvement we want to further improve the safety of all road users. New assistance systems will significantly reduce the hazards for less protected road users such as pedestrians, cyclists or wheelchair users”.
To date, the warning and emergency braking function of production-ready assistance systems is limited to frontal collisions with pedestrians and cyclists in straight-line driving. As part of the project to “protect weaker road users”, existing methods for the image-based recognition of pedestrians were now enhanced to detect as many road users as possible – even when making a turn. This includes pedestrians in an unusual position or stance, partially obscured pedestrians, riders of two-wheeled motorised vehicles, wheelchair users or playing children.
The basis for this are methods developed during the course of the project, which combine empirically determined behaviour patterns with detail information obtained from sensors, such as the direction in which pedestrians look, the orientation of their heads or their leg movements. When merged with sensing of the surroundings and context knowledge, such as the motion of the own vehicle, the course of the lane or the unoccupied space, this allows determining the potential danger more precisely and robustly.
Surroundings model reveals the unseen in the form of a stixel world
As part of the UR:BAN project “surroundings sensing and modelling”, Daimler in cooperation with research partners was able to significantly enhance the performance of surroundings sensing technology in complex urban situations. The researchers now merge the results obtained by stereo cameras and high-performance radar systems to form 360-degree models of the surroundings with a previously unachieved level of detail. The process involves transitioning the some 500,000 three-dimensional pixels surveyed by the stereo camera into an extremely compact representation – the so-called stixel world. This not only contains the position, type, size and motion of other road users, but also stationary obstacles such as parked cars or kerbs. As a result, the unoccupied space for driving is identified reliably and described precisely.
The advance in sensing the surroundings is crucial for urban traffic safety. Because the generated data are of a cross-application nature, they will be able to be used by a large number of widely different systems in the future.This saves costs and resources, thus resulting in a fast propagation of the assistance systems.
Assistance for changing lanes
In addition, Daimler is working intensively on an assisted lane-change function in the “safe lateral and linear guidance in the city” project. Especially changing lanes demands a high level of situational awareness and safe vehicle steering from the driver. Relieving the burden on the driver requires sensing the surroundings and all relevant road users to compute a collision-free lane-change course. The driver activates the lane-change function and always has the possibility to override the system. Initial potential solutions regarding the shape of the trajectory, control concept, time and duration of the lane change could be analysed successfully.
The researchers employed driver monitoring cameras and head-tracking systems to analyse driver intent – the classic look over the shoulder prior to a lane change, for example. These behavioural data are then combined with vehicle data and route information, and used to predict behaviour. The goal is to provide the best possible information to the driver and be able to issue warnings in critical situations as early as possible. A field study on eye movement before and during lane changes found the instrument cluster to be the suitable location to display warnings and assistance messages. These findings will be incorporated into practical applications: the researchers are currently working on a man-machine interface for assisted lane changes.
