The dissertation is divided into three interconnected parts. The first part presents a link travel time estimation algorithm that is based on the use of robust statistic able to exclude the impact of outliers. Outliers in travel time measurements are vehicles whose shortened or extended travel times are not caused by the traffic conditions, but are the result of individual behavior of such vehicle. As the adequate information on travel times is the one of personal cars, the influence of other vehicle categories should be eliminated from the samples which is not feasible with the use of existing link travel time estimation algorithms. The second part of the dissertation presents a method to estimate the value of travel time based on speed extrapolations from point measurements. The method is able to determine whether a speed variation represents a random fluctuation due to individual driver’s behavior and should therefore be smoothed or is a consequence of a change in traffic conditions as a result of a shock wave and should therefore be kept as it is in order to provide prompt response of the algorithm. By extrapolating the speed, the value of travel time from point speed measurements is obtained. In the third part of the dissertation, a data fusion algorithm is presented, combining point and interval detector data to estimate highway travel times also taking into account qualititative measurements of traffic flow. The purpose of travel time data fusion from different sources is to overcome on one hand the spatial inaccuracy of indirect travel time estimation from point speed measurements and on the other hand to overcome the information delay of the direct travel time measurements. By combining both data sources, a short-term travel time prediction is achieved, as the input for the travel time information system.