Vernetzung von Test- und Simulationsmethoden für die Entwicklung von Fahrerassistenzsystemen (Volume 69) (English shop)

Preview

Extract, PDF (990 KB)
Table of Contents, PDF (30 KB)

In the present work, Networking of Test and Simulation Methods for the Development of Driver Assistance Systems, a process for evaluating the attribute-dependent target behaviour of driver assistance systems was developed using ACC as an example, a tool for applying this process in simulation was created, and the applicability of the process in simulation was validated. In a first step (Chapter 2), the state of the art in vehicle system development was compiled in order to obtain an overview of existing tools and development methods and to classify them consistently. Subsequently, driver assistance systems and then the ACC system were presented as examples of modern vehicle systems and finally as the object of investigation (Chapter 3). Building on this compilation, the objectives and approaches of this work were defined (Chapter 4) and the three concrete tasks derived:
• Development of the attribute-dependent target behaviour of driver assistance systems using the example of Adaptive Cruise Control (ACC).
• Conception and construction of an ADAS domain HIL as a simulation tool.
• Validation of the applicability in simulation of the evaluation functions developed through real test drives.
For the first part of the task definition (Chapter 5), the development of the attribute-dependent target behaviour of ACC, an evaluation process was first defined. Thus, defined physical parameters such as velocity or vehicle acceleration must be determined via test scenarios or driving manoeuvres using suitable measurement technology. Since the target behaviour is to be defined as a function of the attributes safety, comfort, dynamics and sportiness, evaluation functions had to be developed with which the physical parameters can be converted into these attributes. For this purpose, the most frequent of the multitude of potentially relevant traffic situations were determined and the following four were selected for further consideration: car following, approaching a queue, overtaking, reaction to a cut-in vehicle and reaction to a cut-out vehicle. By means of a study in which 36 test subjects each had to evaluate three different system configurations, a ranking in the attributes safety, comfort, dynamics and sportiness was determined using a paired comparison and evaluation according to the Law of Comparative Judgement. From a second, identical study with the same test subjects but with a vehicle from a completely different vehicle class (Study I: off-road vehicle; Study II: very sporty mid-range vehicle), the following conclusion could be drawn: for the evaluation of driver assistance systems it can be assumed that the vehicle class or vehicle segment has no influence on the desired target behaviour in the attributes investigated. In addition, the following parameters could be derived from the studies: deceleration and reaction distance for the manoeuvres approaching a queue and reaction to a cut-in vehicle, and acceleration and reaction time for the manoeuvres overtaking and reaction to a cut-out vehicle. From these, a dynamics-reaction plane was constructed, for which evaluation functions could subsequently be developed by combining the subjective and objective study results. A method therefore now exists which makes it possible to make a statement about the probable perception of the system by users on the basis of physical or measurable parameters. The second part of the task (Chapter 6), the stronger integration of simulation into the product development process, was considered in the last section of the work. For this purpose, Hardware in the Loop (HIL) simulation was selected from among the various simulation methods. Accordingly, a concept for an ACC domain HIL that is as modular as possible was developed and implemented. Alongside the components virtual environment, driver and vehicle model, residual bus simulation and simulation environment, an essential element here is the sensor model, since the properties of the respective measurement principle employed, and hence those of the sensor, substantially influence the system configuration. The specific ACC integrated into the HIL uses two radar sensors. For this configuration, the system properties were analysed by means of extensive static as well as dynamic measurements and implemented by suitable methods in the form of a sensor model. Finally, the sensor model was validated. For this purpose, the sensor model was stimulated with reference data from real driving trials and the output of the model was compared with the data of the real sensor system. The evaluation showed comparable results between the model and the real radar sensor. The conclusion was formed by the investigation or comparison of the driving manoeuvres carried out in reality with the simulated driving manoeuvres (Chapter 6.3). For this, the measured variables of the dynamics-reaction plane were determined and compared in each case. Here too the evaluation showed very good agreement, whereby the applicability of the developed evaluation functions in simulation could be demonstrated. In this work, the increased use of simulation in the functional development of highly complex driver assistance systems was investigated through the application of an objective evaluation method. In order to be able to apply the theories developed in concrete terms, the driver assistance system ACC was used as an example. However, a complete integration of the developed methodology, consisting of the evaluation process and the test and simulation tools, into an ACC series development process still requires further work in this field. Thus, the four manoeuvres approaching a queue, overtaking, reaction to a cut-in vehicle and reaction to a cut-out vehicle were considered in the studies and, consequently, evaluation functions were developed only for these. A fully simulated test, with the aid of which developers would be able to make concrete statements, would require an extension of the driving manoeuvre catalogue to include important traffic situations such as driving in dynamic queues, as well as a detailed analysis of the entire area of the dynamics-reaction planes. The transferability of the methodology to other driver assistance systems can be discussed as follows: ACC is a system which supports the driver in his driving task at the guidance level or, owing to the already very high availability and reliability, can practically relieve him entirely of the longitudinal guidance tasks in large parts of the driving task. ACC can thus be counted among the intervening systems with actuators. Further systems belonging to this category at the guidance level are the Lane Keeping Assist (LKA) and the Lane Change Assist (LCA). The informing systems (without actuators) include Lane Departure Warning (LDW), and the autonomous systems with actuators include Automatic Emergency Braking (AEB) or research systems such as the traffic jam pilot. For all these systems, which turn an active vehicle driver into a supervisor of systems, it can be assumed that the methodology developed in the course of this work can also be adapted to them. By contrast, for the systems at the navigation level – navigation system and radio services (RDS-TMC) – as well as systems at the stabilisation level – anti-lock braking system (ABS), traction control system (TCS) and Electronic Stability Program (ESP) – no transferability is expected, or such elaborate modelling of evaluation functions is not expedient. These systems are intended to function in specific situations, according to clear rules and on the basis of easily defined physical measured values, which would thus lead to a simple evaluation such as “test passed” or “test failed”. The developed concept of the ADAS domain HIL as well as the tools and methods applied or developed can also be adapted to the respective system under investigation with only very little effort. In conclusion, it can therefore be stated that this method and above all simulation offers considerable potential for the development of modern vehicle systems at the guidance level.