Dynamic 3D Load Optimization Using AI and Heuristic Integration in Smart Logistics
Downloads
Efficient 3D bin packing remains a significant challenge in logistics, supply chain management, and warehouse automation, where the objective is to maximize space utilization and maintain load stability while minimizing computational time. Traditional heuristic-based methods, such as First Fit and Best Fit, have long been used for their simplicity and speed; however, they often struggle to achieve optimal results in dynamic and complex packing environments. To address these limitations, recent works have explored the use of metaheuristic approaches like Genetic Algorithms (GAs), and more recently, Reinforcement Learning (RL), particularly Proximal Policy Optimization (PPO), to enhance decision-making under constraints. This study proposes a hybrid bin packing solution that combines the strengths of PPO-based reinforcement learning with traditional heuristic strategies to intelligently select item placements in a simulated 3D packing environment. The system was tested using four container sizes and a standardized set of boxes with constraints on volume and weight. Four algorithms—First Fit, Best Fit, Genetic Algorithm, and the proposed Hybrid PPO model—were evaluated using consistent metrics, including packing time, placement success rate, space used, total weight, access efficiency, and stability score. The experimental results reveal that while the First Fit algorithm achieves the fastest packing time (13,269s), it delivers lower placement success (48.4%) and access efficiency (0.60). The Genetic Algorithm achieves high placement rates (52.4–100%) and maximum packing performance, but at a significantly higher computational cost (92,124s). The Hybrid PPO algorithm demonstrates the most balanced performance, achieving a 100% placement success rate in the smallest container and over 72.4% in the largest, while maintaining reasonable packing time (35,712s), high access efficiency (up to 0.95), and superior stability scores (up to 0.80). The Hybrid PPO model outperforms traditional methods and standalone GAs by combining intelligent learning with domain-specific heuristics. This positions the hybrid approach as a promising and scalable solution for real-world logistics environments demanding both efficiency and adaptability in 3D load optimization.
Downloads
[1] Kmiecik, M., & Wierzbicka, A. (2024). Enhancing Smart Cities through Third-Party Logistics: Predicting Delivery Intensity. Smart Cities, 7(1), 541–565. doi:10.3390/smartcities7010022.
[2] Gupta, H., Shreshth, K., Kharub, M., & Kumar, A. (2024). Strategies to overcome challenges to smart sustainable logistics: a Bayesian-based group decision-making approach. Environment, Development and Sustainability, 26(5), 11743–11770. doi:10.1007/s10668-023-03477-6.
[3] Zhou, F., Yu, K., Xie, W., Lyu, J., Zheng, Z., & Zhou, S. (2024). Digital Twin-Enabled Smart Maritime Logistics Management in the Context of Industry 5.0. IEEE Access, 12, 10920–10931. doi:10.1109/ACCESS.2024.3354838.
[4] Saadian, F., Motameni, H., & Golsorkhtabaramiri, M. (2023). Deadline-aware multi-objective IoT services placement optimization in fog environment using parallel FFD-genetic algorithm. Pervasive and Mobile Computing, 92, 101800. doi:10.1016/j.pmcj.2023.101800.
[5] Mesquita, A. C. P., & Sanches, C. A. A. (2024). Air cargo load and route planning in pickup and delivery operations. Expert Systems with Applications, 249, 123711. doi:10.1016/j.eswa.2024.123711.
[6] Lopes, R., Trovati, M., & Pereira, E. (2024). Volumetric Techniques for Product Routing and Loading Optimisation in Industry 4.0: A Review. Future Internet, 16(2), 39. doi:10.3390/fi16020039.
[7] Verma, R., Singhal, A., Khadilkar, H., Basumatary, A., Nayak, S., Singh, H. V., ... & Sinha, R. (2020). A generalized reinforcement learning algorithm for online 3d bin-packing. arXiv preprint, arXiv:2007.00463.
[8] Yang, S., Song, S., Chu, S., Song, R., Cheng, J., Li, Y., & Zhang, W. (2024). Heuristics Integrated Deep Reinforcement Learning for Online 3D Bin Packing. IEEE Transactions on Automation Science and Engineering, 21(1), 939–950. doi:10.1109/TASE.2023.3235742.
[9] Zhang, C., Wu, Y., Ma, Y., Song, W., Le, Z., Cao, Z., & Zhang, J. (2023). A review on learning to solve combinatorial optimisation problems in manufacturing. IET Collaborative Intelligent Manufacturing, 5(1), 12072. doi:10.1049/cim2.12072.
[10] Cai, Q., Hang, W., Mirhoseini, A., Tucker, G., Wang, J., & Wei, W. (2019). Reinforcement learning driven heuristic optimization. arXiv preprint, arXiv:1906.06639.
[11] Zhang, J., & Shuai, T. (2024). Online Three-Dimensional Bin Packing: A DRL Algorithm with the Buffer Zone. Foundations of Computing and Decision Sciences, 49(1), 63–74. doi:10.2478/fcds-2024-0005.
[12] Gao, C., Zheng, Y., Li, N., Li, Y., Qin, Y., Piao, J., Quan, Y., Chang, J., Jin, D., He, X., & Li, Y. (2023). A Survey of Graph Neural Networks for Recommender Systems: Challenges, Methods, and Directions. ACM Transactions on Recommender Systems, 1(1), 1–51. doi:10.1145/3568022.
[13] Spanoudakis, N. I., Akasiadis, C., Iatrakis, G., & Chalkiadakis, G. (2023). Engineering IoT-Based Open MAS for Large-Scale V2G/G2V †. Systems, 11(3), 157. doi:10.3390/systems11030157.
[14] Del Rio, A., Jimenez, D., & Serrano, J. (2024). Comparative Analysis of A3C and PPO Algorithms in Reinforcement Learning: A Survey on General Environments. In IEEE Access. IEEE Access. doi:10.1109/ACCESS.2024.3472473.
[15] Liu, C., & Liu, G. (2024). JointPPO: Diving deeper into the effectiveness of PPO in multi-agent reinforcement learning. arXiv preprint, arXiv:2404.11831.
[16] Kaur, A., & Kaur, B. (2022). Load balancing optimization based on hybrid Heuristic-Metaheuristic techniques in cloud environment. Journal of King Saud University - Computer and Information Sciences, 34(3), 813–824. doi:10.1016/j.jksuci.2019.02.010.
[17] Zhang, W., Maleki, A., & Rosen, M. A. (2019). A heuristic-based approach for optimizing a small independent solar and wind hybrid power scheme incorporating load forecasting. Journal of Cleaner Production, 241, 117920. doi:10.1016/j.jclepro.2019.117920.
[18] Maia, A. M., Ghamri-Doudane, Y., Vieira, D., & Franklin de Castro, M. (2021). An improved multi-objective genetic algorithm with heuristic initialization for service placement and load distribution in edge computing. Computer Networks, 194, 108146. doi:10.1016/j.comnet.2021.108146.
[19] Annie Poornima Princess, G., & Radhamani, A. S. (2021). A Hybrid Meta-Heuristic for Optimal Load Balancing in Cloud Computing. Journal of Grid Computing, 19(2), 21. doi:10.1007/s10723-021-09560-4.
[20] Yan, Y., Liu, G., & Liu, L. (2025). Enhancing Efficiency and Resilience Multiobjective Optimization Model and Algorithm for Site Selection of Emergency Material Storage Sites. KSII Transactions on Internet and Information Systems, 19(2), 344–367. doi:10.3837/tiis.2025.02.001.
- The authors retain all copyrights. It is noticeable that authors will not be forced to sign any copyright transfer agreements.
- This work (including HTML and PDF Files) is licensed under a Creative Commons Attribution 4.0 International License.
