Traffic jams are an economic and environmental problem in many countries. When traffic demand exceeds the freeway capacity, shock waves can be caused by small disturbances in the traffic flow. Since it is not always desirable or feasible to add more lanes on freeways to overcome capacity problems, alternative methods have been developed to reduce and to dissolve traffic jams. One of these methods is using model-based traffic control. The model-based methods use traffic flow models, in which the speeds are typically space mean speeds, while the measured speeds on a freeway are often time mean speeds (measured by loop detectors). The difference between using space mean speeds and time mean speeds for model-based methods has not yet been addressed in literature. In this paper, we investigate the possible performance loss caused by using another mean speed type than the space mean speed for model-based traffic control. Methods for approximating the space mean speed based on local measurements are discussed, together with the time mean speed and geometric mean speed. The suitability of six mean speed types is investigated using microscopic simulation. Next, the three most suitable mean speed types for model-based traffic control are used to determine dynamic speed limits on a freeway using a model predictive control approach. All three types of mean speeds result in the resolution of the congestion in the test scenario and lead to a performance improvement of about 14%.
Samenvatting