Research On Key Technologies For Demand-responsive Transportation Services

Demand-Responsive Transit(DRT)is an emerging operation mode for the public transport sector.Clients are furnished with tailored transport service where the origin,destination,route and length for a single trip are determined by themselves.DRT caters to periodic round trips between urban areas,connects loose traffic between suburbs and rural areas and runs shuttle service for commuter trains and rail transit.In this thesis,DRT as the research object,is devised to achieve transportation accessibility on the demand side and economic viability on the supplier side.Traffic accessibility addresses the ride-matching issue by matching the most suitable supplier for the demander and vice versa.The route planning method,bus stop location method,dynamic bus stop design method of DRT are studied well in this thesis.Economic feasibility seeks to tackle vehicle and crew scheduling to minimize operating costs while meeting the same amount of customer needs.The following work is presented:The theory and key elements of DRT are expounded.The problem definition and concept clarification of DRT are included.The objective function of this thesis is proposed and demonstrated.The key elements such as response time constraint,departure time constraint,time window constraint,demand rejection judgment and design principle of route planning and bus stop location are formulated.A public blockchain designed for DRT service ensures economic viability of the operator by utilizing openness and transparency of the chain to keep travel and consumption of clients on record by intelligently adjusting the charging standard of each road section and the charging in different time periods and different road conditions.This minimizes traffic congestion and improves road efficiency.An experiment that we conducted demonstrates that DRT reduces social vehicle consumption by 92% and 96%,respectively,and social energy consumption by 87% and92%,respectively,compared with ridesharing and noncarpooling.Our team devises a spatial similarity measure for route planning and proposes a comprehensive method based on contraction,insertion,and grid clustering algorithm to accelerate the clustering of trip demands moving in different directions.The contraction-based method can quickly solve this problem.The insertion-based method solves the problem of dissimilarity imbalance and calculates the exact solutions.The dynamic grid-based method is an approximation method based on the divide-and-conquer principle,which narrows the potential search scope,neglects the bad solutions,improves the accuracy of the solutions,and escapes the trap of local optima.The method we design works better with capacitated clustering for large-scale spatial travel by boosting the clustering quality and efficiency.To identify the optimal meeting point(OMP)on road networks,with which clients and facilities reside in non-Euclidean spaces,we propose two efficient heuristic solutions based on approximate and adaptive query processing techniques.These use randomized adaptive search and random direction search methods,respectively,to rapidly converge to the global optimum in the geographic coordinate system.To answer OMP queries effectively,the former adopts a sampling technique in which samples from a larger population are chosen using a method based on the theory of probability to converge to the optimum,and then explores a positive and negative excitation method for area pruning.The latter prejudges the spatial distribution of the OMP by taking advantage of prior knowledge and progressively identifies the optimum step by step from the optimal space to the suboptimal space.An experiment demonstrates that our proposed method enhances the effectiveness by 32.11% compared to the genetic algorithm and simulated annealing algorithm.To optimize ridesharing applications,an integrated ridesharing application was updated from Slugging and Hitchhiking.To be more specific,the hitchhiking form of ridesharing is a way of picking up passengers along the way to accommodate more passengers by increasing intermediate stops and rerouting traffic.This ensures a high ridesharing success rate.In our solution,we analyse the grid cells covered in the route from the origin to the destination and calculate the possible combination of the most popular grid cell(i.e.the most seats occupied).Then,we identify the intermediate stops with the maximum number of passengers served,determine the passenger-to-vehicle assignment,and finally search for the optimal bus stop.Extensive experiments on real datasets demonstrate that the ridesharing success rate of our proposed approaches achieves a 79.63% improvement over the state-of-the-art approach.Mixed integer linear programming(MILP)is proposed for vehicle and crew scheduling in the case of single-and multidepots.We developed a heuristic multiobjective optimization algorithm based on preemptive scheduling.The sum of vehicle and driver costs is typically composed of the fixed cost,variable cost of the vehicle,and duty cost of the driver.However,because there is a trade-off between the cost of dwell time and the cost of idle time,we propose two preemptive scheduling strategies.The former is a preemptive scheduling policy that optimizes objectives by maximizing the total number of vehicles and drivers first and leaving idle cost minimization as a secondary objective.The latter optimizes objectives by minimizing the idle cost first and leaving the total number of vehicles and drivers minimized as a secondary objective.Because the problem is a Bell number problem,the cardinality of the solution space is very large.We propose a multiobjective optimization heuristic algorithm based on preemptive scheduling.The algorithm enumerates all possible combinations and precalculates the costs during the dwell time and the idle time.Then,a random trip sequence is generated by using roulette wheel selection.If the solution is not feasible,then we repair until feasible and sort according to the scheduling policy.Finally,a set partitioning problem is carried out.The total operating cost is diminished by 30.51% and 30.71%,respectively,after applying our algorithm instead of the simulated annealing algorithm and genetic algorithm.