Abstract
Keywords - shifting behavior, measurement system, situation detection, data mining, shifting strategy
Abstract - Automated manual transmissions have been established for a long time now in commercial vehicles. They contribute to an increased driving comfort and to an improved overall safety since they assist the driver in performing routine driving tasks. Despite these benefits, not all drivers favor such automated systems. One reason for this is a discrepancy between the automatic shift patterns and the gears the drivers choose manually, which is again caused by an insufficient detection and interpretation of the drivers intention. Current shifting strategies estimate the drivers intention by observing the accelerator pedal and brake actuation and analyzing vehicle sensor signals and engine torque information. However, traffic situations are so multifarious that the driver intention can not be interpreted entirely by looking only at the information sources mentioned above. Therefore, occasionally unsuitable gears are selected or the shifting is perceived as too hectic or too delayed. To improve automatic shifting a situation detection, a situation related interpretation of the driver intention and situation adaptive shifting strategies are necessary. To realize this, extensive vehicle test drives with commercial vehicles have been performed in a joint research project of DaimlerChrysler and FKFS. On the one hand, their purpose is to investigate the situation related shifting behavior of skilled truck drivers to generate specifications for situation adaptive shifting strategies. On the other hand, the vehicle sensor data was analyzed in conjunction with position data from a Global Navigation Satellite System (GNSS) to develop advanced situation detection algorithms. Therefore, an easy to use measurement system was designed which enables the recording of GNSS data, CAN data and traffic situation describing data on one time basis. The measurement system and the data evaluation process will be presented as well as some results of the investigation of the situation-related shifting behavior. Strategies for an advanced situation detection and their limitations will be discussed. The investigation turned out to be a typical data mining problem. For this reason an outlook on the application of data mining algorithms will be given.