Automated Intersection Volume Counts Using Existing Signal Control Devices
Loading...
Authors
Gholami, Ali
Issue Date
2015
Type
Dissertation
Language
Keywords
ANFIS , Genetic programming , Loop Detector , Mid Intersection Detector , Traffic Signal , Turning Movement Volume
Alternative Title
Abstract
The purpose of this dissertation was to identify and investigate the possibility of obtaining turning volumes from inductive loops and investigate the accuracy of them. A large majority of signalized intersections operate under inductive loops. Experiences in cities such as Seattle, San Antonio, and Toronto show successful usage of inductive loop detectors to obtain traffic volume at intersections. Loop detectors are the most common method for obtaining data at intersections to operate and control traffic signals. In spite of many advantages, they have some drawbacks, including the fact that multiple detectors are usually required to monitor a location.A macroscopic study was performed on two intersections in Reno and Sparks. Both Reno and Sparks use sequential short loops. The detector accuracy was interpreted in terms of count errors. The preferred metric for count error is the Mean Absolute Percent Error (MAPE, %). Results showed the counts were not reliable and had a very high error. At the Kietzke/Moana intersection in Reno, NV, the MAPE was 15 percent northbound, 31 percent southbound, 20 percent eastbound, and 36 percent westbound. At Sparks/Prater in Sparks, NV, the MAPE was worse with all detector groups ranging from 48 to 74 percent.In Reno, advance detector counts could be modified because they showed a strong relationship with base (observed) counts; however, in Sparks, there was not a clear relationship between the two sets of counts. In Chapter 4, by using Genetic Programming (GP) and Adaptive Neuro-Fuzzy Inference System (ANFIS), detector counts were modified and again MAPE was calculated. At Kietzke/Moana, all approaches after data modification had MAPE less than 14 percent. However, at Sparks/Prater, because of the loops’ wiring, there was more irregularity in count detections and as a result, models were not able to reduce detector count errors significantly. Even when detector counts can be modified, detectors are unable to produce turning movement counts in shared lanes. Current practice involves gathering such information through manual counts, which is very costly. Chapter 5 proposes three methods to estimate turning movement proportions in shared lanes. These methods were tested using linear regression and Genetic Programming (GP). It was found that the hourly average error range at intersections was between 4 to 27 percent using linear regression and 1 to 15 percent using GP. The proposed method for modifying detector counts did not guarantee reliable counts in all situations. In Chapter 6, a method is proposed to obtain turning movement counts only from signal information without using detector counts. To produce the required data, a simulation was performed in VISSIM with different input volumes. To change turning volumes, a code was developed in COM interface. With this code, the inputs did not have to be changed manually. In addition, the COM code stored the outputs. Data were then exported to a single Excel file. Afterwards, regression and the Adaptive Neural Fuzzy Inference System (ANFIS) were used to build models to obtain turning volumes. The accuracy of the models was defined in terms of MAPE. Results of the two case studies showed that during peak hours, there was a high correlation between actuated green time and volumes. This method does not require extensive data collection and is relatively easy to employ. The results also showed that ANFIS produced more accurate results compared to regression.Chapter 7 proposes mid-intersection detector (MID) concept configuration to obtain more accurate counts. MIDs are departure doctors which have moved back to middle of intersection. Under this configuration, in addition to stop bar detectors, some mid-intersection detectors also are used to obtain more reliable counts. Due to intersection operation, stop bar detectors were still required, but compared to traditional departure detector configurations, MIDs were expected to produce more reliable and accurate data while requiring same number of detectors.Chapter 8 offers some recommendations to change the loop detector systems for the sake of improving turning movement counts. For obtaining more accurate counts, we recommend: 1) the cost-effective and non-intrusive replacements of inductive loops (Passive Infrared, Active Infrared, Radar and Passive Millimeter, Passive Acoustic, Ultrasonic-Pulse and Doppler). Several “non-intrusive” detection systems are becoming more prominent, being viewed as cost-effective replacements of inductive loops; 2) Changing the configuration and wiring of loops. Performance was significantly enhanced when the loops were connected such that the field generated by the individual loops was additive between the loops rather than subtractive. Counting results were likely to be fair to poor when the loops were separated by 10 or more feet or had a different number of turns or were connected in parallel. To obtain excellent to good counts from loops, each loop should be wired to an individual loop detector channel. If two or more are spliced together into one loop detector channel, the count accuracy would be fair to poor.
Description
Citation
Publisher
License
In Copyright(All Rights Reserved)