基於深度增強式學習之租賃業者資源調度與共享乘車平台合作模式研究;Deep Reinforcement Learning for Resource Allocation and Strategic Collaboration between Rental Agencies and Ride-Sharing Platforms

NCUIR > School of Management at National Central University > Graduate Institute of Information Management > Electronic Thesis & Dissertation > Item 987654321/98252

Please use this identifier to cite or link to this item: https://ir.lib.ncu.edu.tw/handle/987654321/98252

Title:	基於深度增強式學習之租賃業者資源調度與共享乘車平台合作模式研究;Deep Reinforcement Learning for Resource Allocation and Strategic Collaboration between Rental Agencies and Ride-Sharing Platforms
Authors:	葉沛紳;Yeh, Pei-Sheng
Contributors:	資訊管理學系
Keywords:	傳統租賃業;P2P 共享乘車;共享經濟;雙邊市場;競合策略;合作模式;定價策略;增強式學習;深度增強式學習;Traditional car rental industry;Peer-to-peer ride-sharing;Sharing economy;Two-sided market;Coopetition strategy;Cooperation model;Pricing strategy;Reinforcement learning;Deep reinforcement learning
Date:	2025-07-08
Issue Date:	2025-10-17 12:33:04 (UTC+8)
Publisher:	國立中央大學
Abstract:	近年來，點對點（peer-to-peer, P2P）經濟模式的不斷演進，其影響力已滲透至全球各個產業，尤其是在交通運輸領域。P2P 共享乘車平台的興起對傳統租賃公司帶來競爭壓力，迫使部分租賃業者調整經營策略，透過將閒置車輛出租給 P2P 司機來提升車輛使用率，進而增加收益。本研究聚焦於 P2P 共享乘車平台與租賃公司在雙邊市場中的合作模式，探討其相互影響與市場運作方式。本研究設定 P2P 共享乘車平台與租賃公司已採取合作策略。P2P 共享乘車平台主要負責媒合消費者與平台司機，而租賃公司則同時提供車輛租賃服務給一般消費者及無自有車輛卻又想再 P2P 共享乘車平台服務的司機，本研究以雙邊市場合作模式為基礎，從傳統租賃業者角度出發，探討 P2P 平台與租賃公司之間的定價與供給互動機制。首先，建立包含顧客等待反效用、交通混亂程度、自有車輛司機比例及收益分配比例等多項關鍵參數之模擬環境；接著，運用增強式學習（Q-Learning）與深度增強式學習（DQN、DDQN、Dueling DQN、DDPG）等演算法，系統化比較其在不同市場情境下的利潤表現與決策穩定性。最後，基於最佳演算法之敏感度分析，提出動態調整直租價格與車輛供給策略之實務建議，以協助租賃業者在 P2P 競合環境中實現利潤最大化並提升經營韌性。;Recent advances in peer-to-peer (P2P) economic models have transformed industries worldwide, most notably in transportation. The rise of P2P ride-sharing platforms has intensified competition for traditional vehicle rental companies, prompting many to lease their idle fleets to platform drivers in order to boost utilization and revenue. This study examines the cooperative dynamics between P2P ride-sharing platforms and rental companies within a two-sided market framework. We assume an existing partnership in which the platform matches consumers with drivers, while the rental company supplies vehicles both to general customers and to drivers lacking their own cars. From the rental operator’s perspective, we analyze the interplay of pricing and supply decisions. First, we develop a simulation environment incorporating key parameters—consumer waiting disutility, traffic congestion, owned vehicle-driver ratio, and revenue-sharing ratio. We then implement reinforcement learning (Q-Learning) and several deep reinforcement learning algorithms (DQN, DDQN, Dueling DQN, DDPG) to systematically compare their profitability and decision‐making stability across varying market conditions. Finally, drawing on sensitivity analyses of the top‐ performing algorithm, we offer practical guidelines for dynamically adjusting direct- rental rates and vehicle allocations. Our findings aim to equip rental companies with data-driven strategies to maximize profits and strengthen operational resilience in an increasingly P2P-driven transport landscape.
Appears in Collections:	[Graduate Institute of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	20	View/Open

社群 sharing

Loading...