A Methodology to optimize transit road space priority at the network level

Urban traffic congestion is a challenge facing transport networks in almost all world cities. Transit vehicles are efficient at carrying large numbers of passengers in congested road space. This is the major justification for provision of transit priority. Transit priority in the literature can be classified to transit space priority and transit signal priority. While signal priority has been extensively investigated, there are a limited number of studies on transit space priority which deals with the allocation of transit lanes. Previous approaches have a localized focus in nomination of transit lanes. All the proposed methods focus on an evaluation method of some given Transit Priority Alternatives (TPAs); no study has established a comprehensive methodology to optimize transit space priority. The aim of this research is to reallocate the road space between private car and transit modes so that the system is optimized. A bi-level programming framework is adapted to formulate transit space priority. The upper level involves an objective function from the system managers’ perspective; while at the lower level, the users’ perspective is modeled. To take into account the major effects of a priority provision, three models are employed at the lower level: a modal split, a user equilibrium, and a transit assignment model. A set of integer variables define a TPA in the formulation. The outcome of the methodology will determine the optimal TPA. Two approaches are applied to solve the proposed bi-level formulation: Generalized Benders Decomposition (GBD) and Genetic Algorithm (GA). There are a set of constraints that bind the upper level and the lower level. Using the associated dual variables, the GBD approach can decompose the levels. The approach also develops an upper bound and a lower bound on the value of the objective function. In an iterative process, the GBD approach reduces the gap between the bounds. It can be proved that the approach can converge to the optimal answer. Moreover, at each iteration, the GBD approach can determine the gap of the current answer to the optimal point, since it calculates an upper and lower bound. Nevertheless, the approach in some cases may be slow in terms of the computer execution time. The bi-level structure is also solved using the GA approach. The GA is a heuristic approach which can solve the proposed mixed integer non-linear formulation. Although this approach can not guarantee to find the optimal answer, it can achieve a close enough answer to the optimum within a reasonable execution time. The numerical results indicate that the optimal answer was found by the GBD as well as the GA in several small example networks. However, long execution time prevents the GBD to be applied to large scale networks. In contrast, the GA can readily be applied to large networks. Furthermore, considering the execution time of large scale networks, a Parallel Genetic Algorithm is implemented. This algorithm has remarkably reduced the execution time in an example network. It is concluded that the proposed methodology can successfully balance the benefits of all network stakeholders in reallocation of the road space. Two approaches are presented to solve the optimization and the results are compared.