Reinforcement Learning Methods for solving combinatorial optimization problems
| dc.contributor.author | Mustakhov T. | |
| dc.date.accessioned | 2024-12-11T06:23:40Z | |
| dc.date.available | 2024-12-11T06:23:40Z | |
| dc.date.issued | 2023 | |
| dc.description.abstract | The use of reinforcement learning approaches to resolve combinatorial optimization issues in the context of vehicle routing is examined in this paper. Three previously published articles’ insights are combined in one study. The Stochastic Dynamic Vehicle Routing Problem is addressed in the first paper with a deep reinforcement learning strategy. To learn the routing strategy for a single truck as customer orders come in over time, a fully attention-based model with a dynamic encoder and decoder is introduced. The model is taught using reinforcement learning, in which a reward signal influences the choice of nodes to visit. Computational analyses show that this tactic outperforms comparison algorithms. A thorough overview of machine learning techniques used to address NP-hard Vehicle Routing Problems is presented in the second paper. A variety of learning paradigms, solution structures, underlying models, and algorithms are included in the survey. It demonstrates the benefits of machine learning-based models that use the symmetry of VRP solutions and their parity with conventional approaches. The directions for future studies to address the problems with contemporary transportation systems are also described. The third paper provides a thorough analysis of the stochastic dynamic vehicle routing problem, examining state-of-the-art approaches and methodologies for solving this challenging optimization problem. The stochastic and dynamic restrictions that present significant challenges to efficient route planning and optimization are covered in detail. The study offers a thorough grasp of cutting-edge techniques and recommends prospective directions for further investigation. In addition, the use of Markov decision processes and sophisticated reinforcement learning techniques to address the issue of stochastic dynamic vehicle routing is examined. Combining these articles, this thesis advances knowledge of reinforcement learning methods for combinatorial optimization issues, particularly in the area of vehicle routing. The study demonstrates the effectiveness of deep reinforcement learning and machine learning techniques while also highlighting areas that need additional study and development in the field of transportation logistics. | |
| dc.identifier.uri | https://repository.sdu.edu.kz/handle/123456789/1567 | |
| dc.language.iso | en | |
| dc.subject | address NP, learning paradigms, vehicle routing, transportation logistics | |
| dc.title | Reinforcement Learning Methods for solving combinatorial optimization problems | |
| dc.type | Other |