Yield sign behavior

All runs for this paper were first done with an experimental code and then repeated with the Transims production code; all results shown so far were obtained from the Transims production code. The disadvantage of an experimental code is that actual implementation in the production version may still introduce changes in the results due to small discrepancies.^32.13 The advantage of an experimental code is that turnover (compile times, complexity of code, etc.) is much better than with a production version. We used that advantage to test many different rules. In the following, we want to present a small subset of tests.

All results presented in this section refer to the situation of a 1-lane minor street merging into a 1-lane major street, with the intersection control being a yield sign. Fig. 32.6 (a) shows what happens if the ``reservation'' rule from the Transims production code is no longer used. Clearly, if vehicles from the major road do reserve cells on the outgoing link only if they are actually going there, many more vehicles from the minor lane can make the turn, effectively leading to an ``alternating'' vehicle pattern. This may be desirable in some situations.

Figs. 32.6 (b) shows what happens when one then changes ``accept when $gap \ge 3 v_{oncoming}$'' to ``accept when $gap > 3 v_{oncoming}$''. This seems like a negligible difference in the rules; yet, the results are quite different in the congested regime. Whereas in the first, many vehicles are able to get into the congested major road, in the second, only few of them make it. The difference is easiest explained by looking at a vehicle of speed zero on the major road just in front of the merge point, with space for a vehicle downstream of the merge point. With the first rule, a vehicle at the yield sign will accept the move and move in front of the vehicle on the major road, in the second case, it will not. Both scenarios seem to be plausible to us; only systematic measurements can probably resolve which one is better for a simulation model. - Also note that the rule in (b) generates similar flows as the Transims production version.

Fig. 32.6 (b), (c) and (d) show the result of different speed limits (same speed limit for both streets). A high average free speed of approx. 130 km/h ($\approx 80$ mph, generated by $v_{max}=5$), maybe a freeway merge, generates a flow of approx. 2000 veh/hour/lane in the incoming lane when there is no traffic on the major road (Fig. 32.6 (c)). From there, maximum incoming flow decreases continuously. Lower average free speeds of approx. 75 km/h (50 mph, Fig. 32.6 (b)) and 50 km/h (30 mph, Fig. 32.6 (d)) generate lower maximum incoming flows and are generally closer to the Highway Capacity Manual curve. Yet, it should be clear that, contrary to the HCM, the ``minor'' flow is also a function of the speed limit and not only of the gap acceptance (the gap acceptance is the same in all three simulations).

A last series of experiments shows the effect of different values for the gap acceptance. Figs. 32.6 (e) and (f) show ``accept when $gap > v_{oncoming}$ and $gap > v_{max}$''. Clearly, more vehicles are accepted, leading to a higher flow of turning vehicles as a function of the flow on the major road. Note that the flow via the yield sign is never higher than 1800 minus the flow on the major road. This reflects the fact that the major road cannot have a higher flow than 1800 veh/h/lane (free speed approx 50 mph); traffic through the yield sign can thus at most fill the major road to capacity. This explains why the acceptance of much smaller gaps do not produce a stronger difference. The situation is clearly different for unprotected turns across instead of into traffic, as can be seen for the left turns in the next section.