 |
International Society of Science and Applied Technologies |
| Agent-Based Analysis of Network Reliability in Stochastic Flow Networks for E-Commerce Logistics | ||||
| Author |
Wen-Hui Kuo
|
|||
| Co-Author(s) |
Ping-Chen Chang
|
|||
| Abstract | Existing studies on reliability evaluation in stochastic-flow networks typically prescribe a fixed routing policy in advance under a single operating scenario. However, real-world e-commerce logistics involves continuously changing demand levels and delivery deadlines, making a single static routing rule insufficient. This study proposes an agent-based simulation framework comprising three types of agents. ArcAgents simulate stochastic arc capacities by sampling from empirical distributions, constituting the underlying network environment. Independent CarrierAgents are each assigned to a distinct minimal path. Each learns accept-or-reject decisions through tabular Q-learning, using only the bottleneck capacity of its assigned path as the decision state, without requiring knowledge of arc capacity distributions or centralized coordination. A RouterAgent serves as a platform-level coordinator, dynamically adjusting carrier priority based on empirical delivery performance to manage order assignment. The results show that the proposed framework achieves high system reliability across varying operating conditions without exhaustive policy enumeration. This demonstrates that decentralized multi-agent learning is an effective approach for routing reliability analysis in e-commerce logistics systems.
|
|||
| Keywords | stochastic-flow network (SFN), agent-based simulation, Q-learning, system reliability, multi-agent decision-making | |||
| Article #: RQD2026-16 | ||||
Proceedings of 31st ISSAT International Conference on Reliability & Quality in Design |