On motorways and expressways Variable Message Signs (VMS) are installed to warn drivers if potentially dangerous situations may exist. The situations are estimated based on measurements of the environment. The measurements are conducted by sensors, but the quality of the sensors is currently not supervised sufficiently within the road network of the ASFINAG. Erroneous sensors are detected either late or just randomly.
This project will produce a prototype which compares sensor data systematically in order to detect potential errors within the sensor measurements. The availability of sensor and the quality of each sensor data will be evaluated. A “sensor availability index” will be generated for all sensors using a series of heuristics. Single values will be checked prior and after data transmission. The newly developed software component will utilize the existing relational database of sensor data installed and maintained by the ASFINAG: Additionally the sensor data will be analysed with statistical and automatic checking methods known from the field of "Knowledge Discovery in Databases (KDD)“. The KDD checks plausibility and generates a “Performance Index”. These methods are based on probabilistic algorithms extracted from the area of data mining and artificial intelligence/machine-based learning.
The results of the analysis are probabilistic variables. Both the “performance index” and the “availability index” are stored in real time in the prototype. The prototype will require separate measurements to check the validity of the sensor data. Therefore, separate measurements will be done in a test area for reference. Two sites will be selected, and additional sensors will be installed for specifically selected sensor types. The measurement data will be integrated in the data base for reference purposes. A software client for service and maintenance will be developed in implemented for sensor data configuration. Automatically sensor quality reports will be generated.