We present a two-echelon van-based robot last-mile pickup and delivery system in urban settings. Robots can visit areas with van access restrictions, such as pedestrianized areas or university campuses. The van stops at parking nodes to drop and/or pick up its robot, and to replenish its robot and/or swap its robot's battery if needed. Five van/robot pickup and delivery cases are considered, according to the roles the van/robot plays in the process of pickup/delivery and whether the van transports its robot. To model the proposed problem, we introduce a mixed-integer program including time, freight, and energy. We further propose an adaptive large neighborhood search algorithm to solve larger instances and a capacity feasibility test approach for a single route. We then assess the infuence of parking node density on model output. A case study based on a realistic city scenario is introduced. A sensitivity analysis is performed on the robot's travel cost and maximum travel distances, and the e ect of van no-go areas is studied. Two classical models (two-echelon vehicle routing with hybrid pickup and delivery and parallel van and robot scheduling with hybrid pickup and delivery) are compared with ours, and results show our model is competitive in appropriate scenario settings. We therefore advocate using the two-echelon van-based robot last-mile pickup and delivery system in urban areas.