详情

原文传递 SIMULATION FOR PLANNED FIELD TEST OF NEURAL NETWORK CONTROL OF TRAFFIC SIGNAL TIMING.

题名：	SIMULATION FOR PLANNED FIELD TEST OF NEURAL NETWORK CONTROL OF TRAFFIC SIGNAL TIMING.
作者：	Chin-Daniel-C; Smith-Richard-H
关键词：	adaptive-control-systems; computer-simulation; real-time-control; traffic-control; traffic-signals
摘要：	This paper reports on the progress in the preparation for the field test of the system-wide traffic-adaptive control (S-TRAC) technique for traffic signal timing. S-TRAC uses a closed-loop neural network to control traffic signal timing in real time and the daily operation information to optimize the neural network weights over days and weeks. The Department of Transportation of Montgomery County Maryland has kindly offered their facility as an S-TRAC test site. We have completed most preliminary analysis for the field test, and working toward the field demonstration. The preliminary analysis includes site selection, measure-of-effectiveness selection simulation study, and a preliminary traffic impact study. The impact study is the only task not yet completed. The simulation study uses the real data taken from the south Rockville area that includes 11 intersections consisting of two major intersected arteries. The traffic arrival rates are varied every 5 minutes according to the field data over two morning rush hours during the period 7:00 to 9:00 a.m. A total of fifty simulated cases shows an average 7-percent increase of traffic counts compared to the original fixed timing control. The 90-percent confidence bound is within 2.5 percent. The increases are mainly due to fewer cars waiting for the traffic signals at the end of the control period. In other words, the late arriving cars do not need to wait long for the early arrivals to leave to system.
总页数：	ITS America. Meeting (7th : 1997 : Washington, D.C..). Merging the transportation and communications revolutions : conference proceedings. 1997. pp11
报告类型：	科技报告