A large number of accidents on the road can be led back to the lack of knowledge of the current road conditions and thus the coefficient of friction between the road and tires. Often, the driver can only react with great delay to changes in the coefficient of friction and adjust his driving style. The same applies to active driver assistance systems, such as the emergency brake assist, which, if friction coefficient is too low, only reduce the impact speed but can not prevent the accident. The goal of this project is to estimate the road condition through probability-based data fusion of different heterogeneous sources and to transfer it into a time-varying map, the Road Condition Cloud (RoCCl). This information can be used by the vehicles, for example, to warn the driver of slippery roads early or to adapt the driver assistance systems accordingly. Within each vehicle, a detailed vehicle dynamics model with the aid of series sensors such as accelerations, yaw rate and wheel speeds is used to estimate the current friction coefficient of the tire-road contact and additional confidence in appreciating the accuracy or correctness of the estimation. In addition, information from onboard and environmental sensors (temperature, windscreen wiper status, GPS position, distance behavior, etc.) will be consulted. This information is sent to the RoCCl and processed there in real-time. In this case, data of a very large amount must be taken into account at high data rates and a large variety of data. Probability-based sensor data fusion methods are used to process the heterogeneous information of individual vehicles in one location into an estimation of the road condition, again with additional confidence. By adding global information such as temperature history or precipitation information, the individual estimations are extended into a global, time-varying map. Thus, the confidence of the estimation at a particular location can be increased by information from multiple vehicles. In addition, there is the possibility of interpolating locations with an estimated value that are not directly stimulated by information from a vehicle. The communication with the RoCCl also offers the possibility to initialize driver assistance systems at the beginning of a journey correctly. Temporal abrupt changes from dry to wet due to incoming rain or slow changes due to drying are also taken into account.The functionality of the RoCCl will be demonstrated in a simulation with a scalable fleet size. Within the simulated fleet several real vehicles are integrated to validate the results on real test vehicles.

A large number of accidents on the road can be led back to the lack of knowledge of the current road conditions and thus the coefficient of friction between the road and tires. Often, the driver can only react with great delay to changes in the coefficient of friction and adjust his driving style. The same applies to active driver assistance systems, such as the emergency brake assist, which, if friction coefficient is too low, only reduce the impact speed but can not prevent the accident. The goal of this project is to estimate the road condition through probability-based data fusion of different heterogeneous sources and to transfer it into a time-varying map, the Road Condition Cloud (RoCCl). This information can be used by the vehicles, for example, to warn the driver of slippery roads early or to adapt the driver assistance systems accordingly. Within each vehicle, a detailed vehicle dynamics model with the aid of series sensors such as accelerations, yaw rate and wheel speeds is used to estimate the current friction coefficient of the tire-road contact and additional confidence in appreciating the accuracy or correctness of the estimation. In addition, information from onboard and environmental sensors (temperature, windscreen wiper status, GPS position, distance behavior, etc.) will be consulted. This information is sent to the RoCCl and processed there in real-time. In this case, data of a very large amount must be taken into account at high data rates and a large variety of data. Probability-based sensor data fusion methods are used to process the heterogeneous information of individual vehicles in one location into an estimation of the road condition, again with additional confidence. By adding global information such as temperature history or precipitation information, the individual estimations are extended into a global, time-varying map. Thus, the confidence of the estimation at a particular location can be increased by information from multiple vehicles. In addition, there is the possibility of interpolating locations with an estimated value that are not directly stimulated by information from a vehicle. The communication with the RoCCl also offers the possibility to initialize driver assistance systems at the beginning of a journey correctly. Temporal abrupt changes from dry to wet due to incoming rain or slow changes due to drying are also taken into account.The functionality of the RoCCl will be demonstrated in a simulation with a scalable fleet size. Within the simulated fleet several real vehicles are integrated to validate the results on real test vehicles.