Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network
In last-mile delivery, drivers frequently deviate from planned delivery routes because of their tacit knowledge of the road and curbside infrastructure, customer availability, and other characteristics of the respective service areas. Hence, the actual stop sequences chosen by an experienced human d...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier BV
2024
|
| Online Access: | https://hdl.handle.net/1721.1/156448 |
| _version_ | 1811069742297907200 |
|---|---|
| author | Mo, Baichuan Wang, Qingyi Guo, Xiaotong Winkenbach, Matthias Zhao, Jinhua |
| author_facet | Mo, Baichuan Wang, Qingyi Guo, Xiaotong Winkenbach, Matthias Zhao, Jinhua |
| author_sort | Mo, Baichuan |
| collection | MIT |
| description | In last-mile delivery, drivers frequently deviate from planned delivery routes because of their tacit knowledge of the road and curbside infrastructure, customer availability, and other characteristics of the respective service areas. Hence, the actual stop sequences chosen by an experienced human driver may be potentially preferable to the theoretical shortest-distance routing under real-life operational conditions. Thus, being able to predict the actual stop sequence that a human driver would follow can help to improve route planning in last-mile delivery. This paper proposes a pair-wise attention-based pointer neural network for this prediction task using drivers’ historical delivery trajectory data. In addition to the commonly used encoder–decoder architecture for sequence-to-sequence prediction, we propose a new attention mechanism based on an alternative specific neural network to capture the local pair-wise information for each pair of stops. To further capture the global efficiency of the route, we propose a new iterative sequence generation algorithm that is used after model training to identify the first stop of a route that yields the lowest operational cost. Results from an extensive case study on real operational data from Amazon’s last-mile delivery operations in the US show that our proposed method can significantly outperform traditional optimization-based approaches and other machine learning methods (such as the Long Short-Term Memory encoder–decoder and the original pointer network) in finding stop sequences that are closer to high-quality routes executed by experienced drivers in the field. Compared to benchmark models, the proposed model can increase the average prediction accuracy of the first four stops from around 0.229 to 0.312, and reduce the disparity between the predicted route and the actual route by around 15%. |
| first_indexed | 2024-09-23T08:15:15Z |
| format | Article |
| id | mit-1721.1/156448 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T08:15:15Z |
| publishDate | 2024 |
| publisher | Elsevier BV |
| record_format | dspace |
| spelling | mit-1721.1/1564482024-09-01T03:59:55Z Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network Mo, Baichuan Wang, Qingyi Guo, Xiaotong Winkenbach, Matthias Zhao, Jinhua In last-mile delivery, drivers frequently deviate from planned delivery routes because of their tacit knowledge of the road and curbside infrastructure, customer availability, and other characteristics of the respective service areas. Hence, the actual stop sequences chosen by an experienced human driver may be potentially preferable to the theoretical shortest-distance routing under real-life operational conditions. Thus, being able to predict the actual stop sequence that a human driver would follow can help to improve route planning in last-mile delivery. This paper proposes a pair-wise attention-based pointer neural network for this prediction task using drivers’ historical delivery trajectory data. In addition to the commonly used encoder–decoder architecture for sequence-to-sequence prediction, we propose a new attention mechanism based on an alternative specific neural network to capture the local pair-wise information for each pair of stops. To further capture the global efficiency of the route, we propose a new iterative sequence generation algorithm that is used after model training to identify the first stop of a route that yields the lowest operational cost. Results from an extensive case study on real operational data from Amazon’s last-mile delivery operations in the US show that our proposed method can significantly outperform traditional optimization-based approaches and other machine learning methods (such as the Long Short-Term Memory encoder–decoder and the original pointer network) in finding stop sequences that are closer to high-quality routes executed by experienced drivers in the field. Compared to benchmark models, the proposed model can increase the average prediction accuracy of the first four stops from around 0.229 to 0.312, and reduce the disparity between the predicted route and the actual route by around 15%. 2024-08-29T15:22:44Z 2024-08-29T15:22:44Z 2023-07 2024-08-29T15:19:16Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/156448 Mo, Baichuan, Wang, Qingyi, Guo, Xiaotong, Winkenbach, Matthias and Zhao, Jinhua. 2023. "Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network." Transportation Research Part E: Logistics and Transportation Review, 175. en 10.1016/j.tre.2023.103168 Transportation Research Part E: Logistics and Transportation Review Creative Commons Attribution-Noncommercial-ShareAlike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Elsevier BV arxiv |
| spellingShingle | Mo, Baichuan Wang, Qingyi Guo, Xiaotong Winkenbach, Matthias Zhao, Jinhua Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network |
| title | Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network |
| title_full | Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network |
| title_fullStr | Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network |
| title_full_unstemmed | Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network |
| title_short | Predicting drivers’ route trajectories in last-mile delivery using a pair-wise attention-based pointer neural network |
| title_sort | predicting drivers route trajectories in last mile delivery using a pair wise attention based pointer neural network |
| url | https://hdl.handle.net/1721.1/156448 |
| work_keys_str_mv | AT mobaichuan predictingdriversroutetrajectoriesinlastmiledeliveryusingapairwiseattentionbasedpointerneuralnetwork AT wangqingyi predictingdriversroutetrajectoriesinlastmiledeliveryusingapairwiseattentionbasedpointerneuralnetwork AT guoxiaotong predictingdriversroutetrajectoriesinlastmiledeliveryusingapairwiseattentionbasedpointerneuralnetwork AT winkenbachmatthias predictingdriversroutetrajectoriesinlastmiledeliveryusingapairwiseattentionbasedpointerneuralnetwork AT zhaojinhua predictingdriversroutetrajectoriesinlastmiledeliveryusingapairwiseattentionbasedpointerneuralnetwork |