實驗賽局探討多重資訊服務提供者(ISP)之路徑導引策略

Abstract

Information Service Provider (ISP) is an important role in the Advanced Traffic Information System (ATIS) who provide traffic related information to the subscribed road users or travelers. The information provision strategy is therefore quite critical in that same information regarding traffic condition and guidance direction may actually result in collecting too many users merging toward same road section at the same or very short period time and create unwanted congestion. Being aware of such possible dilemma, this thesis try to initiate an exploration of game-like behavior among ISPs to provide suitable information for route guidance under such uncertain traffic outcomes which may be due to the collective effect of various ISPs’ strategies. A three-stage study was conducted. In the first stage, pre-game questionnaire was surveyed to recognize the basic characteristics of domestic ISPs such as the market share and technical capability to enable a categorization process. In addition, data including the attitude of ISPs in generalizing their information strategies were included. A series designed scenarios were also employed to examine the possibilities of game-like actions. A day-to-day dynamic controlled simulation experiment was implemented in the second stage to explore in depth the corresponding information and particularly guidance strategy by those selected ISPs being as players in an interactive multi-player game under various scenarios according to an experiment design with information fullness and completeness as two major controlled factors. The experimental area is of the Northern region of Taiwan, with main studied OD pair where from Taipei to Shintsu with about 6 major routes identified. Observations throughout the ten-week experiment of how various ISPs (five ISPs in total) offer their route guidance for users at different time intervals for users located at different locations during peak hours were analyzed to reveal the pattern of information provision and the resulting payoffs by each individual ISP and as a whole. Finally, a decision model framework based on indifference band concept was developed at the last stage of this study to capture the evolution of route guidance decisions over day and over time period within the same day respectively. Each model was specified in a time-to-time manner under which three models were calibrated accordingly to match three major periods (lasting several weeks respectively) representing different scenarios of game elements according to experiment design. Results of estimation showed very high degree of goodness-of-fit overall, and almost all parameters were with statistical significance including those associated with game-related variables. These results were in line of the hypothesis that each individual ISP may or may not switch its route guidance evolutionarily over time for considering the presence of other ISPs and revealed information.