S-Space College of Engineering/Engineering Practice School (공과대학/대학원) Dept. of Civil & Environmental Engineering (건설환경공학부) Theses (Ph.D. / Sc.D._건설환경공학부)
Determining Depot Locations and Capacities for Demand Responsive Autonomous Vehicles : 수요응답형 자율주행차의 Depot 입지 및 용량 결정
- 공과대학 건설환경공학부
- Issue Date
- 서울대학교 대학원
- Demand responsive autonomous vehicle (DRAV) ; Location and capacity of depot ; Bi-level model ; Genetic algorithm (GA)
- 학위논문 (박사)-- 서울대학교 대학원 : 공과대학 건설환경공학부, 2018. 2. 고승영.
- In the automobile and transportation sector, various technologies related to an autonomous vehicle (AV) and shared mobility are expected to upgrade. Accordingly, various types of service using AV have been proposed recently in the field of Demand Responsive Transport (DRT) such as a shared autonomous vehicle (SAV), which is one-way car sharing services using AV. The Demand Responsive Autonomous Vehicle (DRAV) system is a system that utilizes AV for on-demand service in the public domain. It can improve the travel conveniences for users in blind spot of transport service areas and vulnerable road users (e.g. disabled, elderly). Additionally, since this system also is used as a demand management policy by promoting the sharing transport, DRAV can be considered as a competitive new para-transit option. However, although many previous studies are concerned about road congestion possibly caused by a large number of AV, lack of studied considered such problem. DRAV depot not only decreases the road congestion but also plays multiple roles such as effective vehicle management and charging infrastructure of future rechargeable battery vehicles.
This study aims to develop a model and an algorithm to determine optimal location, quantity, and capacity of DRAV depot considering road congestion due to empty AV travels after the introduction of the system. Iterative modal split and traffic assignment procedures are deployed in the model to describe users behavior more realistically. Moreover, a solution algorithm based on genetic algorithm (GA), a representative meta-heuristic technique, is developed to solve the NP-hard combinatorial optimal solution problem type in a reasonable time. The characteristic of the problem that the solution pattern varies according to the number of depots is considered in the algorithm. EMME4, a typical transportation simulation program, and python 2.7 are utilized for efficient analysis and the problem-solving process is automated by using application programming interface (API).
The Mandl's network is selected for a case study analysis. Results reveals that network congestion cost due to empty vehicle travel should be considered in the DRAV depot decision process. Depot location is determined by high DRAV demand, but the additional construction of depot is recommended for highly congested areas to reduce the congestion cost. Further scenario analyses, represents the various future situations, proves that influencing factors for the depot location selections such as local transportation environment and location-related factors (e.g, traffic volume, travel pattern, public transport route, land cost) for location selections should be fully considered.