The rapid growth of e-commerce has led to a surge in package deliveries, particularly during promotional events, resulting in increased workload for deliveryman. To address this challenge, this paper explores traveling salesman problem with bike and two autonomous delivery robots-referred to as 2TSPBR. The robots, each with three different-sized cargo compartments, are responsible for delivering packages of varying sizes. When their inventory is low, they restock from the bicycle’s service points. The Variable Neighborhood Search (VNS) algorithm is employed to generate and optimize delivery routes for the bicycle and robots. The results indicate that the VNS algorithm significantly outperforms the CPLEX solver in terms of computational time, providing rapid convergence to near-optimal solutions. This study demonstrates the efficiency and effectiveness of the proposed VNS algorithm in delivering accurate solutions for the integrated delivery system.