Synchronizing last trains of Urban Rail Transit (URT) system is important for passenger transportation, especially for travelers from other modes, such as high-speed railway (HSR) trains in late night, to reach their destinations. This paper develops a bi-objective mixed-integer linear programming model for the problem of Last-Train Timetable Synchronization (LTTS) of URT system in late night hours. The bi-objective optimization model maximizes the number of passengers who can arrive at their destinations successfully, and minimizes the total operation ending time for last trains of URT system. This model can deliver both a last-train transfer scheme solution (i.e. which transfer connections to keep at which stations) and a last-train timetable of URT system (i.e. arrival and departure times for last trains at stations) at the same time. Transfer walking time is considered different for passengers, and suggestions for passengers to adopt certain transfer connections are dynamically dependent on the last-train timetable and passenger assignment. The Epsilon-constraint method is used to find the Pareto optimal solutions for the proposed bi-objective model. Numerical tests based on a sample network demonstrate the validity of the model on solving the LTTS problem. To evaluate the effectiveness and efficiency of the bi-objective model in practice, real-world experiments are conducted based on Beijing-Shanghai high-speed railway (HSR-BS) line, Beijing-Tianjin inter-city high-speed railway (HSR-BT) line and Beijing urban rail transit (URT) system. We also present a set of experiments to evaluate the benefits of considering different Transfer Walking Time (TWT) of passengers in the model, compared to models with constant and identical TWT for all passengers.