Optimale verkeerslichtenregelingen met prioriteit voor openbaar vervoer

Traffic signals and queues at intersections are major components of public transportation (transit) delay. Active transit signal priority (TSP) strategies are used as an effective way of reducing that delay. They can be divided in green split adjustment and phase resequencing measures. However, the effectiveness of TSP strategies is dependent on the traffic conditions, the frequency of priority requests, and their specific impact on the signal control. Some strategies can add significant delay to non-priority flows. From the transit operator’s point of view, this becomes problematic when an intersection contains several conflicting transit lines in mixed traffic conditions. Serving priority for one phase can lead to substantial additional delay for a transit line served by another phase. In networks, TSP strategies could cause spillbacks to other intersections serving transit lines. From the road operator’s point of view, spillbacks of non-priority flows are mostly not desired. The key feature of a proper multimodal optimization of the signal settings is to take conflicting objectives into account in a balanced manner. In this research, models and insights are delivered for optimizing signal settings (i.e. offsets, green splits and cycle lengths) when TSP strategies are applied. The focus is on active TSP strategies such as phase extension (i.e. green extension), phase advance (i.e. red truncation), and phase resequencing strategies like rotation of the transit phase for early service, insertion of an extra transit phase, and skipping of non-priority phases. The thesis can be divided in three research parts: 1) Development and validation of a new stochastic traffic and optimization model CAPACITEL for optimizing green splits, cycle lengths of isolated intersections (chapter 3) and extra optimization dimension of offsets of corridors with TSP (chapter 4). 2) Development and validation of analytical formulas for optimal signal settings of isolated intersections with transit signal priority using CAPACITEL (chapter 6 and chapter 5); 3) Comparison of the effectiveness of various active TSP strategies for isolated intersections (chapter 7 and chapter 8) and in combination with coordination recovery strategies for arterials (chapter 9). In the first part, new signal setting formulas are proposed for optimizing the green splits and cycle lengths. The scope of these equations is isolated intersections when TSP strategies are applied, applicable for undersaturated, near-saturated and oversaturated traffic demands. Webster’s optimal green split and cycle length formulas are still considered relevant in many signal timing manuals. However, TSP strategies can influence the optimal green splits and cycle lengths of isolated intersections. These effects are now researched for adapting the Webster formulas accordingly. For updating the Webster formulas for a multimodal environment, signal settings are optimized for various combinations of the following parameters: intersection flow ratio, transit and non-transit phase flow ratio, transit frequency, detector location, lost time, and the type of TSP strategy. The coefficients of the new formulas are then calibrated with those results by a linear regression analysis. Two categories of formulas of minimal delay are identified: 1) TSP strategies with limited capacity shift like phase rotation in combination with green extension, but without red truncation (chapter 5); 2) TSP strategies with red truncation (chapter 6); TSP strategies with red truncation significantly change green splits. A new optimal green split formula is developed that is applicable for 2, 3 and 4-phase intersections. The critical lane method – as proposed by Webster – is used for calculating the green splits for TSP strategies without red truncation. The main reason is that a TSP request impacts the green split significantly when red truncation is applied. When milder TSP strategies are used, the green split is hardly influenced when considering a full hour of simulation. For both categories of TSP, a new optimal cycle length formula is suggested. The regression formula for the TSP strategies with red truncation has 29 parameters and correlations of those parameters. The number of parameters and their correlations in the formula is limited to 3 for the strategies without red truncation. The latter formula is thus more suitable for practitioners and easier to use. Both categories of formulas have a significantly better performance in minimizing delay compared to existing alternatives and are best applicable for near-saturated and oversaturated traffic demands. In the second part of the research, various TSP strategies are compared on their effectiveness in reducing person or vehicle delay. A dynamic TSP control with an early call detection and a confirmation detection is developed and compared with full priority in chapter 7. For comparing their effectiveness, a sensitivity analysis is performed for different intersection flow ratios, weights for the performance function, location of the detector and impact of a transit stop. Dynamic TSP is preferred in undersaturated conditions because of its flexibility of redistributing green lengths without going to oversaturation. When the intersection gets saturated, the favour shifts to full priority with a short distance detector (150m). Transit stops shift the favour towards a TSP strategy with a detector beyond the stop. Twelve TSP strategies are compared for their effectiveness in chapter 8. The impact of these TSP strategies is investigated for undersaturated, near-saturated and oversaturated conditions for a wide range of transit and traffic demand scenarios. In order to guarantee a fair comparison, the basic signal settings are optimized and tailored to each TSP strategy and demand so that the total person delay at the intersection is properly minimized. This is in contrast with chapter 7 where TSP strategies were applied after the optimization of the signal settings. Phase skipping and strategies that apply red truncation are not effective in reducing person delay. They are mostly effective in reducing bus delay but they come at a cost of adding person delay to other users. Green extension, phase insertion, phase insertion with green extension and phase rotation with green extension appear to be the most effective TSP strategies in minimizing intersection person delay. For undersaturated conditions the choice of the TSP strategy has a limited effect on the objective function and travel time reliability. Once an intersection reaches near-saturated or undersaturated conditions, strategies with red truncation or phase skipping shouldn’t be used. In the last part, the performance of a new stochastic mesomodel for corridors, CAPACITEL, is validated. CAPACITEL incorporates microscopic arrivals and departures, but models spillbacks and queues macroscopically. By using less detail for the queuing, it is able to perform simulations a lot faster than VISSIM. That makes the model more suitable for optimization purposes. On the other hand, CAPACITEL is detailed enough to simulate buses and grasp the stochastic impact of TSP strategies on the signal settings and the corresponding flows. Besides the aforementioned TSP strategies, a fixed cycle length and a green compensation strategy is available for coordination recovery after a transit vehicle is served. In this way local – i.e. serving TSP – and global – i.e. arterial progression – multimodal objectives can be combined. The optimization of the signal settings in CAPACITEL is split in three subproblems. Green splits, cycle lengths and offsets are optimized in series with the new multimodal formulas of chapter 5 as a starting point. When optimizing the network, the optimization of each subproblem starts at the heaviest loaded intersection. Neighbouring intersections are optimized one by one starting from the heaviest loaded intersection till the end of the arterial. CAPACITEL demonstrates its superiority for optimizing the signal settings of multimodal arterials by outperforming TRANSYT15 for a test case in Zeebrugge, Belgium. In conclusion, a decision tree is developed where all the insights from this research are incorporated. This decision tree proposes a methodology for designing optimal multimodal signal settings.status: publishe