PRORETA5

Enabling Automated Driving in Urban Traffic

The goal of the interdisciplinary and interuniversity research project PRORETA 5 is the investigation of artificial intelligence methods with focus on situation understanding and motion planning for enabling automated driving in urban traffic.

For this purpose, the Institute of Automotive Engineering (FZD) in cooperation with Continental, the Control Methods and Robotics Lab at TU Darmstadt, the Institute of Cognitive Neuroinformatics at University of Bremen as well as the Faculty of Automatic Control and Computer Engineering at TU Iaşi (Romania) is researching automated driving in urban environments.

Here, the focus is on automated, predictive driving using artificial intelligence. In particular, it aims to gain an understanding of driver actions and environment data, traffic rules and the interaction between road users. By imitating driver behavior, humanized motion planning will be implemented and an understanding of driver attention in relation to external objects and situations will be achieved. Based on this, feature reduction and compression methods will be implemented and complex maneuvers, adaptive driving under environmental influences and cooperative driving strategies will be realized. A systems engineering methodology will be developed for the integration and testing of the AI.

PRORETA 5 is the fifth season of the successful research cooperation between Continental and TU Darmstadt, which is based on the results and experiences of the previous projects PRORETA 1 (emergency brake and evasion assistance), PRORETA 2 (overtaking assistance), PRORETA 3 (integral safety concept and cooperative automation) and PRORETA 4 (safety by learning), and which was supplemented in PRORETA 5 by two further universities as research partners.

In automated driving, research is being conducted on the use of artificial intelligence for various modules such as environment detection, behavior prediction of other road users or trajectory planning. Since these novel methods are also accompanied by unknown uncertainties, which can lead to dangerous trajectories, a module for safety verification of the planned trajectory is provided.

This module contains several submodules that use only rule-based methods, i.e. without artificial intelligence. This includes the plausibility check of information from the environment sensor system as well as internal vehicle sensor data, such as lidar detections or wheel speeds. In further submodules, the functionality of the overall system, the stability of the trajectory planner and the collision criticality of other road users to the planned trajectory are analyzed. If one of these submodules indicates that the planned trajectory cannot be executed safely, an alternative trajectory, such as an emergency stop trajectory, is sent to the controller instead. In this way, safe vehicle behavior is to be ensured at all times.

This topic is being researched by Christoph Popp (opens in new tab).

The provision of a safety case for highly automated vehicles remains unresolved. One challenge is the specification of test cases and test evaluation criteria. The aim of this research work is to develop a description of the required behavior of an automated vehicle in an urban environment in order to derive test concepts including pass/fail criteria. The main focus is on testing the conformity with traffic regulations. Here, the influence of relevant aspects of the environment on the behavior of an automated vehicle is investigated. This leads to a behavior-based functional specification of an automated vehicle for urban road traffic. By means of a hazard and risk analysis, safety goals and requirements for the system are identified and broken down to module level. These requirements are transformed into a test concept for a prototype vehicle.

This topic is being researched by Felix Glatzki (opens in new tab).