EXPERIMENTAL STUDY OF AN ADAPTIVE ROUTING ALGORITHM FOR C2C LOGISTICS

Abstract

The article presents the results of an empirical study of an innovative adaptive routing algorithm developed to enhance the efficiency of parcel delivery systems in the customer-to-customer (C2C) logistics segment. This approach addresses the critical “last-mile” delivery challenge — the final stage of transporting goods directly to end-users — which is increasingly relevant due to the rapid growth of online marketplaces and mobile trading platforms. The proposed algorithm combines Long Short-Term Memory (LSTM) networks for demand forecasting with Reinforcement Learning (RL) techniques for real-time decision-making.

Experimental testing was conducted in a mid-sized city environment, representing a typical urban ecosystem with heterogeneous transportation infrastructure. Real-world data on traffic conditions, weather patterns, and delivery demand dynamics were used in the simulations. The results demonstrate a substantial reduction in logistics costs (up to 18%) and delivery times (by 15–20%) compared to traditional methods such as Dijkstra and A* algorithms. Furthermore, the algorithm showed a high level of adaptability to changing conditions, making it suitable for integration into dynamic C2C logistics platforms.

The paper details the research methodology, including the structure of the testing framework, evaluation metrics (cost, time, accuracy), and result validation procedures. Particular attention is given to a comparative analysis with existing approaches, the identification of limitations, and prospects for further development. The study is relevant both to academic research and to businesses focused on developing flexible, scalable next-generation delivery systems.