In general, traffic’s reaction to influential factors is not linear. One possible approach to a better way of interpreting and predicting traffic conduct is the use of artificial intelligences (ai), which have already been implemented before in several scientific papers regarding traffic analysis. The Mobility Department was able to turn this method into a practicable solution in order to replace classical evaluation algorithms and to improve the quality of the analyses. The research starts, where these traditional methods are either overstrained because of non-linear correlation and a multitude of parameters, can only provide inaccurate results or react too sensitively on external interferences.
An important parameter for traffic analysis is the determination of up-to-date travel times on certain routes, although the basis for data varies. Mainly, raw information on the route to be examined is provided by FCD but also stationary, locally measuring traffic sensors.
If the travel time cannot be determined sufficiently, information on traffic density or a classified status report will be given to the traffic participant alternatively. But this information has to be comprehensible to all recipients – also considered that each one has a different perspective on traffic situations.
Additional to the determination of the actual state, it is also necessary to further develop and test means of traffic prognosis – ranging from predictions for the next minutes as well as for several hours.
Because different routes show different traffic related conditions (e.g. edificial, judicial, etc.) analytical parameters have to be adapted individually. With means of ai-methods, this adaptation can be performed automatically.
peter.maurer@ait.ac.at