AKAMATSU Takashi Nomura Research Institute Researcher, 研究員
MIAYAKI Yasuyuki Asikagao inst. of technology Asso. Prof., 土木工学科, 助教授 (90174135)
MATSUMOTO Yoshiji Tokyo Science Univ. Faculty of Eng. Prof., 理工学部, 教授 (20010663)
SANO Kazushi Tokyo Univ. Faculty of Eng. Assistant, 工学部, 助手 (00215881)
IEDA Hitoshi Tokyo Univ. Faculty of Eng. Asso. Prof., 工学部, 助教授 (90168089)
桑原 雅夫 東京大学, 生産技術研究所, 助教授 (50183322)
宮本 和明 横浜国立大学, 工学部, 助教授 (90150284)
|Budget Amount *help
¥4,000,000 (Direct Cost : ¥4,000,000)
Fiscal Year 1990 : ¥1,500,000 (Direct Cost : ¥1,500,000)
Fiscal Year 1989 : ¥2,500,000 (Direct Cost : ¥2,500,000)
Transportation network equilibrium assignment model which originates from the one step in conventional transportation demand analysis has grown to one of the important model in various area, e. g. transportation planning, land use planning, traffic management etc., with the developments of integrated demand analysis model or generalized equilibrium model. When we apply this model to the real transportation network, however, the model often includes numerous number of variables and enormous calculational costs are required. To reduce the computational burden, various method, for example, Frank-Wolfe method, projection gradient method, sub-gradient method etc., are proposed. The computational times and costs required in these methods, however, are not yet easy to dealt with large scale network for the practical wide area analysis. As the result, we often see the situations that these transportation models which are developed as the tool for giving planners qualitative and accurate information are not fully utilized.
So far as we premise to use the existing serial processing computer, radical improvement of this problem may be difficult. Considering the fact that all the variables used in in the network assignment model have similar expression and structure, however, it may be possible to solve the problem very rapidly by parallel calculation, assuming that there are some parallel processing method based on proper principle.