Seite - 35 - in Maximum Tire-Road Friction Coefficient Estimation

2 Estimation of the friction potential Both the high amount of required knowledge of vehicle and tire parameter and the signal processing effort can be partially circumvented by using ANN that identify tem- poral patterns within input and output structures, which requires training the networks beforeapplication. Lex et al. showedthat frictionestimationusing recurrent neuralnet- works (RNN) was accurate enough to adapt the intervention strategy of an automated emergency brake assist, [LKE13a], [LKE13b]. Nevertheless, the results were not always replicable. Thus, even for similar driving states, different estimates were found. In addi- tion, neural networks are not suitable for extrapolating to conditions they have not been trained for. This means that it would be necessary to conduct extensive training that considersall relevantdrivingandroadconditionswitharealisticprobabilitydistribution. Rajamani et al. showed three different approaches for estimatingµmax based on three different sensor configurations, [RPPL12]. In addition, different models were used de- pending on the sensor configuration, which focused either more on the vehicle or the wheel level. An algorithm based on the wheel motion showed better convergence and estimation accuracy than an algorithm based on a vehicle model. These results are consistent with the results of the sensitivity analysis presented in Section 4. 2.2.3. Car-to-x communication systems (C2x) In contrast to pure on-board solutions, the idea of C2x approaches is to combine traffic and road-related information about other traffic participants (C2C) and roadside in- frastructure systems (C2I). A variety of information (e.g. on traffic density, accidents ahead, road conditions of specific road sections) can be transmitted wirelessly to the vehicle. Currently, the standardisation of data transfer, including data format and con- tent, is being addressed. In 2010, the European Commission issued a mandate to the standardisation organisations CEN2, CENELEC3 and ETSI4 to develop standards for cooperative systems in intelligent traffic systems, [EotEC09]. The data will be trans- ferred via the standard IEEE802.11, which has been especially developed for wireless ad hocnetworks, [Ins10]. Oneexampleofapure infrastructure-basedsystemis theWeather Data Management System (WDWS), where information from local sensors on the street (e.g. humidity, road surface temperature, precipitation) is used in combination with large-area weather data (e.g. from weather radar) to determine the actual road condi- tion and to calculate a prognosis. This system is already being used to support road 2European Committee for Standardization 3European Committee for Electrotechnical Standardization 4European Telecommunications Standards Institute 35