Experiment with conditional bus-priority at controlled intersections in Eindhoven

In the city of Eindhoven (the Netherlands), all vehicles of the public transport company (Hermes) have an on board computer that computes the punctuality deviations constantly (even while between stops) and requests priority at controlled intersections only when it determines that the vehicle is late. This system of conditional priority reduces punctuality deviations a lot, as can be seen by comparing the graphs below.

The first graph was produced on a day that the on board computers did not have a valid timetable, thus no priority was requested by the on board computers.
For the second graph the on-board computers used a timetable that was created from observations made with absolute priority for the buses at all controlled intersections. This timetable was computed using TRITAPT.

The schedule was the same for both graphs.

In the graphs below, the fine green lines indicate punctuality deviations of individual vehicles. The bold blue lines are the 15% and 85% values, while the bold black line indicates the mean punctuality deviation.
The stops are shown along the horizontal axis, the punctuality deviations along the vertical axis. If a bus is exactly on time along the whole route, there should be a green horizontal line at level 0. If a bus is late, the green line is above 0, if a bus is early below 0.

Punctuality deviations with no priority. In this case the buses got no special treatment at controlled intersections.

Punctuality deviations with conditional priority. The buses were given priority only when the on-board computer indicated that the the bus was late. If an on-board computer (for any reason) did not have the correct timetable loaded, no priority was given.

The first graph clearly indicates that punctuality deviations vary widely, though the majority of vehicles was late. Drivers could not do much to correct that situation.
The second case (which is the normal state of affairs in Eindhoven) shows what can be achieved with conditional priority. The lines spread out only slightly because getting early with respect to the timetable is almost impossible during the first half of the route (there are no controlled intersections on the second half of the route). Evidently, there were a few trips that started more than 4 minutes late, or more than one minute early. These buses were apparently "adjusted" by the conditional priority system.