Pickup and delivery problems are a classic routing problem. This paper attempts to solve a one-to-many-to-one pickup and delivery problem with time windows by introducing a R2 indicator based multi-objective evolutionary algorithm. In the formulation, three conflicting objectives, i.e., the number of vehicles, the travel distance, and the total workload are optimized in an evolutionary context. Furthermore, a R2-based local search is imposed to repair and tune the offspring for a better validity. The proposed algorithm is evaluated on a PDP-TW benchmark dataset.