Demand-responsive bus scheduling optimisation considering candidate pick-up and drop-off points

Demand-responsive transit systems have emerged as a vital solution for enhancing public transportation efficiency; however, they suffer from inefficiencies in operations. Implementing flexible origin–destination solutions can help minimise travel time, thereby enhancing overall operational performance. The unrestricted choice of pick-up and drop-off points introduces substantial complexity in vehicle scheduling and routing optimisation. Thus, this paper introduces an adaptive large neighbourhood search (ALNS) algorithm to optimise the bus scheduling process by dynamically responding to demand fluctuations and optimising candidate pick-up/drop-off points. Roulette destroy and best repair operators are applied with random operators for getting out of local optima. Simulated annealing acceptance is activated when the best repair operator is used. Experimental validation on the Wangjing road network demonstrates that incorporating candidate points leads to a significant improvement of 11% in system efficiency. In addition, the proposed destroy and repair method upon ALNS algorithm further improves efficiency by an 17%. In sensitivity analysis, candidate point demands are capable of making extra service possible in traffic jams and saving more time compared to traditional cases. These results validate the proposed approach and highlight the potential of the adaptive algorithm in addressing the challenges of demand-responsive bus scheduling.