In urban metro systems, train delays make planned timetables uncoordinated, especially during the end-of-service period. Therefore, this study focuses on the timetable rescheduling problem to improve network accessibility. A multi-objective programming model is first proposed to optimize the headway between the last and penultimate train and station dwell time. Minimization of the passenger transfer waiting time for all origin-destination (OD) pairs is meaningful to improve passenger transport efficiency, and a penalty time is used instead of the transfer waiting time of inaccessible passengers, which is important to decrease inaccessible passenger volume. On the other hand, minimization of the total last-train dwell time for all lines is beneficial for decreasing the operating costs for metro operation companies. Then, the adaptive non-dominated sorting genetic algorithm-II (ANSGA-II) is combined with a label setting method to obtain the Pareto optimal solutions. Finally, the Nanjing metro network with randomly-designed delay scenarios is adopted to verify the effectiveness of the proposed rescheduling model and solution algorithm. The results show that this model can yield rescheduling schemes to address the different sudden delay events, resulting in a significant decrease in passenger transfer waiting time and last-train dwell time, thereby improving the network accessibility.INDEX TERMS Adaptive non-dominated sorting genetic algorithm-II, delay management, label setting method, last train, network accessibility.