Beitrag psychologischer Modelle und Methoden zur Bewertung von Fahrerassistenzsystemen.

Der Beitrag diskutiert Problembereiche, Gestaltungsaufgaben und psychologische Erkenntnisse bei Entwurf und Konstruktion von Fahrerassistenzsystemen (FAS). Bislang existiert kein Standardverfahren, um Nutzen und Risiko neuer FAS identifizieren und bewerten zu können. In einer beispielhaften Anwendung der Fahraufgabenmethodik SAFE werden Anforderungen an den Fahrer verglichen, je nachdem, ob sie mit oder ohne FAS bewältigt werden müssen und wie sich die FAS-Komponenten auf Komplexität und Risiko einzelner Teilaufgaben auswirken. Effekte ergaben sich nicht nur in einer verdichteten Bewertung, sondern ließen sich auch in einzelnen Anforderungskategorien nachweisen. Somit erscheint eine Methodik wie SAFE auch gut für solche Zwecke einsetzbar.

Contribution of psychological knowledge and methods to the assessment of ADAS

Safety problems in car-driving are easily associated with driver errors. Therefore, engineers have called for more automation to replace human functions by machines in man-machine systems in the past. But experience in human factors engineering has shown that automation can have disadvantages like deskilling or overconfidence on the side of system users. In car-driving, the idea of assistance has substituted pure automation. The driver is to be supported by information, warnings or interventions, but he can always override the system, thus still being part of the control loop. Assistance systems are nowadays available at all three levels of the driving task. Their development leads to changed task allocations between the driver and the car/system compared to conventional cars. The introduction of the contribution derives somes inferences from this fact. At first, possible problems associated with the intoduction of ADAS are mentioned and shortly discussed. A focus of the discussion is the ergonomic design of coded information. The information flow between the driver and the car and its subsystems should be on time, restricted to relevant data, situation specific , acceptable and clearcut. Because of the need for situational adequacy of the information presented, the designer should have deeply understood the central role of the driving task in the conception and design of assistance systems.

The authors have shortly constructed a procedure for analysing driving task requirements. It is based on a model of human information processing that is used to model the driver and a taxonomy of driving tasks. The text gives a short summary of the central ideas of the methods and its main procedural steps.

The main part of the paper describes the application of the method to evaluate the safety impact of a planned assistance system. The system is composed of a navigation aid and a function to support lane-changing. The driving task under investigation is the approach to a signalised junction including a lane change to the left to prepare a left-turning manoeuvre in the junction. The idea is to model this task using the SAFE-method twice: with and without the assistance system and to compare thr resulting requirements for the driver with respect to complexity and risk in the different subtasks.

The results show that the assistance system makes the task easier for the driver. Parts of the navigation subtasks are assessed as low complex ones compared to a rating of medium complexity without the system. The early announcement of the junction by the system helps to gain time, thus reducing time-pressure linked to various subtasks in the navigation context. Through the redundancy of information in the speech output component, display elements and out of the vehicle information, the accuracy of information intake by the driver is of less importance. The sum of mental requirements is lower with the assistance system. There are less memory and decision processes nesessary, reducing the ratio of conscious to automatic processes. In summary, the complexity of the navigation subtasks in the approach to the intersection is lowered by the system components and the navigation descision is supported in a powerful way.

The estimation of the risks point into the same direction, based on lower time-pressure with the system. Moreover, less errors are expected with the system. Typical lane change errors like “changing to a wrong lane” or “lane-changing too late” are less probable.

After the navigation task has been completed the lane-changing task itself has to be managed. The subtasks related to that phase of the junction approach also gain from the system messages. The ratings of complexity tend to be lower compared to the case without system. Time - pressure and required accuracy of some mental requirements are less critical, e.g. because the system supports difficult judgements (time gap to car on the the left lane too short). Several demands on driver expectation and memory are eased as well. Whereas the complexity of the subtasks related to the lane-changing manouevre is lowered considerably, its riskiness does not change in the two cases.

At the end of the contribution some conclusions were drawn. In the framework of conception and design of ADAS the driver remains the central component of the driver-car-environment system. A reasonable application of an ADAS presupposes a detailed knowledge of the requirements of each driving subtask. The information need of the driver has not only to be determined with respect to its quality, extent and temporal structure, but also the information exchange between the system components has to be optimized to bridge the gulf between the biological system (driver) and the physical system (car).

Aus: Fastenmeier, W. & Gstalter, H. (2008). Beitrag psychologischer Modelle und Methoden zur Bewertung von Fahrerassistenzsystemen. Themenhef "Ergonomie im Fahrzeug“, Zeitschrift für Arbeitswissenschaft, 62, 15-24.

zurück zur Übersicht