| Project/Area Number |
17560471
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
交通工学・国土計画
|
|---|
| Research Institution | Shinshu University |
|---|
Principal Investigator |
OKUTANI Iwao Shinshu University, Engineering Dept., Professor, 工学部, 教授 (90026138)
|
|---|
| Project Period (FY) |
2005 – 2006
|
| Project Status |
Completed (Fiscal Year 2006)
|
| Budget Amount *help |
¥1,700,000 (Direct Cost: ¥1,700,000)
Fiscal Year 2006: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 2005: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | Optical beacon / Traffic signals control / Wavelet neuron / Neural network |
|---|
| Research Abstract |
In this study we first accomplish learning models of neuron-type, i.e., the three layered wavelet neuron model and neural network model which can estimate travel time over a link between two adjacent intersections from the inputs made up of signal split at upstream as well as down stream intersection, offset, traffic volumes on the upstream feeding links in addition to the subject link, the link length. The well learned models are then employed to optimize the traffic signals control parameters such as green signal times and offset. The proposed method is tested over an arterial street with five signalized intersections standing in a line by making use of the widely accepted traffic simulation tool called NETSIM. The optimized control pattern is compared to several non-optimal control patterns over the time period of 19 signal cycles (cycle duration equals 150sec.). It is exhibited that the signal control pattern optimized through the proposed method consistently brings about less total travel time than those generated under other control strategies provided for comparison.
Finally, it should be mentioned that the proposed method yields stable solution when the input value of traffic volume fluctuates in accordance with normal distribution with 20% coefficient of variation since the resulting offsets differ from optimum values only up to two seconds, with no difference in green time solution.
|
