The demand for last-mile delivery is increasing due to the rise of ecommerce and urbanization trends, leading to various social problems. To address these problems, the introduction of autonomous delivery robots is being researched. However, most stud...
The demand for last-mile delivery is increasing due to the rise of ecommerce and urbanization trends, leading to various social problems. To address these problems, the introduction of autonomous delivery robots is being researched. However, most studies on delivery robots for the last mile sector assume customer attendance delivery, so it is essential to develop robots capable of door-to-door delivery. This paper proposes a 24-hour delivery operation system for urban environments. The system uses door-to-door delivery robots and applies a combinatorial optimization strategy based on deep reinforcement learning using algorithms such as STRUCT2VEC, LSTM and ATTENTION. This enables the system to solve routing problems effectively and in a short time to meet customer requirements.