| spelling |
oapen-20.500.12657-890582024-04-03T02:23:00Z Chapter Scheduling Optimization of Electric Ready Mixed Concrete Vehicles Using an Improved Model-Based Reinforcement Learning Chen, Zhengyi Zhang, Xiao Song, Changhao Cheng, Jack C. P. Electric vehicle Ready-mixed concrete delivery Scheduling optimization Model-based reinforcement learning Monte Carlo Tree Search thema EDItEUR::U Computing and Information Technology Decarbonizing the construction sector has become an imperative global agenda, with electric machinery playing a pivotal role in realizing this objective. This research concentrates on devising an operational scheduling optimization method for electric ready-mixed concrete vehicles (ERVs) – a groundbreaking, eco-friendly intervention for the construction sector. We commence by outlining a systematic problem definition for the ERV operational process, considering the distinctive characteristics of electric vehicles and ready-mixed concrete (RMC) delivery tasks. The entire process is then conceptualized as a Markov decision problem (MDP), which enables sequential decision-making. We subsequently develop an enhanced model-based reinforcement learning technique, named parallel-masked-decaying Monte Carlo Tree Search (PMD-MCTS), for efficient resolution of the MDP. The entire system is authenticated via a real-world case study, and the PMD-MCTS's performance is juxtaposed against existing benchmarks. The results demonstrate the appropriateness of the proposed MDP formulation for tackling RMC delivery tasks. The PMD-MCTS algorithm and one of its ablation algorithms (PM-MCTS) have demonstrated superior performance compared to other benchmarks in either cost reduction or delay minimization, with PMD-MCTS requiring 30% less computation time than PM-MCTS 2024-04-02T15:45:10Z 2024-04-02T15:45:10Z 2023 chapter ONIX_20240402_9791221502893_27 2704-5846 9791221502893 https://library.oapen.org/handle/20.500.12657/89058 eng Proceedings e report application/pdf n/a 9791221502893_74.pdf https://books.fupress.com/doi/capitoli/979-12-215-0289-3_74 Firenze University Press 10.36253/979-12-215-0289-3.74 10.36253/979-12-215-0289-3.74 bf65d21a-78e5-4ba2-983a-dbfa90962870 9791221502893 137 12 Florence open access
|
| description |
Decarbonizing the construction sector has become an imperative global agenda, with electric machinery playing a pivotal role in realizing this objective. This research concentrates on devising an operational scheduling optimization method for electric ready-mixed concrete vehicles (ERVs) – a groundbreaking, eco-friendly intervention for the construction sector. We commence by outlining a systematic problem definition for the ERV operational process, considering the distinctive characteristics of electric vehicles and ready-mixed concrete (RMC) delivery tasks. The entire process is then conceptualized as a Markov decision problem (MDP), which enables sequential decision-making. We subsequently develop an enhanced model-based reinforcement learning technique, named parallel-masked-decaying Monte Carlo Tree Search (PMD-MCTS), for efficient resolution of the MDP. The entire system is authenticated via a real-world case study, and the PMD-MCTS's performance is juxtaposed against existing benchmarks. The results demonstrate the appropriateness of the proposed MDP formulation for tackling RMC delivery tasks. The PMD-MCTS algorithm and one of its ablation algorithms (PM-MCTS) have demonstrated superior performance compared to other benchmarks in either cost reduction or delay minimization, with PMD-MCTS requiring 30% less computation time than PM-MCTS
|
