Physical Internet (PI) was introduced to transform the current logistics systems into a global sustainable logistics network. This paper focuses on two-way scheduling/grouping in both sections of a Road-Rail PI-hub (Road→Rail and Rail→ Road). It consists in scheduling both inbound and outbound trucks as well as grouping of PI-containers in the train’s wagons and outbound trucks. The problem is formulated as a mixed integer programming model (MILP). Then, a simulation-optimization (SO) approach is proposed to provide robust solutions which remain feasible in spite of perturbations. The proposed approach uses a Modified Threshold Accepting meta-heuristic (MTA) as an optimizer combined with an embedded perturbation simulator to ensure the robustness of the solutions. Results show that the proposed SO approach provides robust solutions able to handle perturbations.