A feasibility study of dilemma zone problems, performed by collecting and analysing traffic flow data at a high-speed signalised intersection, showed that the maximum green extension or cutback needed to get a vehicle out of the dilemma zone is generally no more than 2 seconds. If we scan all vehicles on a link just a few seconds before the beginning of a yellow interval, we may be able to extend or cut back the green interval so that the vehicles can avoid the dilemma zone. For each vehicle approaching an intersection on the link that is about to turn yellow, there is a time interval such that (a) the vehicle will be in the dilemma zone without green extension (or cutback) or (b) it will be in the dilemma zone when there are T seconds of extension (or cutback). So the task is to find the smallest nonnegative integer T that is not in any of these time intervals and extend (or cut back) the current interval by T seconds. This T can always be found and assuming there are no other restrictions, dilemma zones can be avoided and the extension (or cutback) is done at most once for each green interval. The simulation study, performed by modifying the source codes of NETSIM, showed that the signal timing generated by a bandwidth maximisation program (PASSER-II) resulted in a lower number of vehicles in the dilemma zone than that generated by a delay minimisation program (TRANSYT-7F). Additionally, the signal timing generated by the combination of the two programs, that is, by minimising delay within the constraint of bandwidth maximisation, resulted in an even lower number of vehicles in the dilemma zone than those generated by each program alone. The technique developed in this study can be implemented if the speeds and positions of all vehicles on the roadway can be recorded at small time intervals (e.g., 1 sec). Recommendations are made for implementing and testing the developed technique.
Abstract