EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1340. The dynamic travelling salesman problem with time-dependent and stochastic travel times: a deep reinforcement learning approach
Invited abstract in session TB-28: Advancements of OR-analytics in statistics, machine learning and data science 5, stream Advancements of OR-analytics in statistics, machine learning and data science.
Tuesday, 10:30-12:00Room: 065 (building: 208)
Authors (first author is the speaker)
| 1. | Dawei Chen
|
| 2. | Christina Imdahl
|
| Eindhoven University of Technology | |
| 3. | David Lai
|
| University of Southampton Business School | |
| 4. | Tom van Woensel
|
| Technische Universiteit Eindhoven |
Abstract
We introduce a novel approach using deep reinforcement learning to tackle the Dynamic Traveling Salesman Problem with Time-Dependent and Stochastic travel times (DTSP-TDS). The main goal is to dynamically plan the route with the shortest tour duration and visit all customers while considering the stochastic nature of travel times. We employ a reinforcement learning approach to dynamically address the stochastic travel times to observe changing states and make decisions accordingly. Our reinforcement learning approach incorporates a Dynamic Graph Temporal Attention model that dynamically extracts information from the stochastic environment. We have tested the performance of our proposed approach on the simulation. Our approach can quickly provide high-quality solutions for all datasets using limited computing resources. The efficiency study on different models demonstrates that new components, such as the selection mechanism, dynamic attention component, and temporal pointer can improve the DGTA model. These findings highlight the importance of incorporating stochastic elements in the model to achieve better performance. Furthermore, the trained DGTA model exhibits generalization capability to different instances with varying customer locations and uncertainty levels of travel times without requiring additional training time. Our work contributes to advancing the field of DTSP-TDS, with potential applications in various industries, such as healthcare logistics.
Keywords
- Machine Learning
- Logistics
Status: accepted
Back to the list of papers