Optimization analysis of the number and location of holding control steps : A simulation-based evaluation of line number 1, Stockholm

Abstract: Summary   The growing congestion problems in big cities result in growing need for public transport services. In order to attract new users, public transport operators are looking for methods to improve  their performance  and level of service.   Service reliability  is one of the main   objectives   of  public  transport   operators.   Various   sources   of  service   uncertainty   can   causebus  bunching:  buses  from  the same  line tend  to bunch  together  due to a positive feedback  loop,  unless  control  measures   are  implemented.   The  most  commonly   used strategy for preventing service irregularity is  to define holding points along the bus route. The design of the holding strategy involves the determination  of the optimal number and location of holding points, as well as the holding criteria. These strategies are classified to schedule-  or  headway-based.   Previous  studies  showed  that  headway-based   strategies have the potential  to improve  transit  performance  from both passengers  and operators perspectives.   This thesis analyzes the performance of optimization algorithms when solving the holding problem. The optimization process involves the determination  of time point location for a given  headway-based   strategy.   The  evaluation   of  candidate   solutions   is  based  on  a mesoscopic  transit simulation.  The input data for the simulation  corresponds  to the bus line number 1 in Stockholm city.   The  objective  function  is  made  up  of  the  weighted  sum  of  all  time  components  that passengers  experience:  in-vehicle  riding  time,  dwell  time,  waiting  time  at stop  and  on- board holding time.The optimization was carried out by greedy and genetic algorithms.  In addition, a multi-objective  function that incorporated  the performance  from the operator perspective was solved using a multi-objective genetic algorithm.   The results demonstrate  the potential benefits from optimizing the location of time point stops.  The  best  solution  results  in  an  improvement   of  around  11%  in  the  objective function value. Interestingly, the results indicate that wrongly chosen time point stops can yield transit performance that is worse off than having no holding control.