A Multi-objective Evolutionary Algorithm for Eco-efficient DRTS Problems

論文翻譯標題: 應用多目標演化式演算法求解Eco-efficient需求反應式車隊派遣問題
  • 林 婉婷

學生論文: Master's Thesis


Environmental impact has become an important issue since the serious disasters related to climate change have happened frequently worldwide in recent years Based on the White Book on Transportation Policy Taiwan (MOTC 2012) the transportation sector accounts for 12 9% of the total national energy use and generates about 13 9% of the national CO2 emissions In particular the road sector has accounted over 90% CO2 emissions of transportation sector; therefore related research on CO2 emissions reduction by applying advanced transportation technologies increase recently Demand responsive transit services (DRTS) aims at providing efficient vehicle routes to satisfy customer requirements under limited resources In order to reflect environmental cost in DRTS a formulation of the multi-objective dial-a-ride problem with the consideration of eco-efficiency (fuel consumption and emission) is developed and a multi-objective evolutionary algorithm is constructed to find the Pareto optimal solution Three objectives specifically considered in the research include customer disutility system cost and environmental cost The first two objectives represent total travel time and cost of vehicles The environmental cost is evaluated through emission which is represented as a function of travel speed Unlike single objective modeled problem multi-objective approach helps to generate a compromised solution with no predefined parameters or weights It can not only generate a more realistic and accurate solution but also makes route scheduler to make decision easier with more information about the trade-off between different objectives Compare to ?-Constraint method multi-objective evolutionary algorithm found the Pareto front more efficient by using a population approach to generate more than one solution in each run For this reason it can generate globally optimized solutions which are not sensitive to the shape of Pareto front The multi-objective evolutionary algorithm is solved through a revised genetic algorithm with non-dominated sorting techniques Numerical experiments are conducted to justify the proposed approach and to evaluate the quality of Pareto solution The experiments are conducted based on a real city network by using a simulation-assignment model DynaTAIWAN
獎項日期2014 九月 1
監督員Ta-Yin Hu (Supervisor)