Compared to using connected vehicle data alone, algorithms developed to estimate the positions of unequipped vehicles can result in up to an 8 percent reduction in delay.

Real-Time Prediction of Vehicle Locations in a Connected Vehicle Environment.

Date Posted
09/25/2017
Identifier
2017-B01181
TwitterLinkedInFacebook

Real-Time Prediction of Vehicle Locations in a Connected Vehicle Environment

Summary Information

The purpose of this study by the Virginia Center for Transportation Innovation and Research was to develop and investigate techniques to estimate the positions of unequipped vehicles based on the behaviors of equipped vehicles in a connected vehicle environment. This study was undertaken in the hope that these techniques, by providing sophisticated estimates of unequipped vehicle locations, could improve the performance of other proposed connected vehicle mobility applications at low penetration rates with minimal associated costs.

Methodology
Four tasks were carried out to achieve the study objectives:

Literature review
The literature was reviewed to determine similar proposed techniques to estimate positions of individual unequipped vehicles in a connected vehicle environment at low penetration rates. Additional literature was reviewed to allow a better understanding of car following models and connected vehicle technology and to identify connected vehicle mobility applications that could benefit from position estimates of individual unequipped vehicles.

Development of freeway and arterial location estimation algorithms
Two algorithms were developed to estimate the positions of unequipped vehicles based on the behaviors of equipped vehicles. One algorithm was developed for freeway environments (where vehicles react primarily to other vehicles) and the other for arterial environments (where vehicles often queue at traffic signals). The algorithms were designed to be as generic as possible, with minimal calibration to specific road networks.

Development of testing parameters for the developed algorithms
The next phase of the research plan developed the parameters that would be used in testing the algorithms (e.g. the evaluation network, number of simulation repetitions, and connected vehicle application). Others were identified during the testing task, such as sensitivity of an estimated vehicle’s lifespan on freeways.

Testing and evaluation of the developed algorithms
The developed algorithms were tested extensively. The freeway location estimation algorithm was tested using vehicle trajectory data collected from the Next Generation Simulation (NGSIM) project that consisted of a 500-meter (1,640-foot) section of I-80 in Emeryville, California. The arterial location estimation algorithm was tested on a calibrated model of U.S. 50, a four-signal arterial network in Chantilly, Virginia.

In all tested scenarios, multiple simulation runs were evaluated with unique random seeds for various penetration rates, ranging from 5 percent to 100 percent. Each algorithm was applied to an existing connected vehicle mobility application (signal control for the arterial location estimation algorithm, and ramp metering for the freeway location estimation algorithm) to determine if the location estimation algorithms developed could improve the performance of these applications at low EV penetration rates.

Findings
Two connected vehicle mobility applications were able to use these estimates to produce small performance improvements in simulation at low penetration rates of connected vehicle technologies when compared to using connected vehicle data alone

  • Ramp metering application (freeway location estimation algorithm): up to an 8 percent reduction in delay was achieved.
  • Traffic signal control application (arterial location estimation algorithm): up to 4.4 percent reduction in delay was achieved.
Goal Areas
Results Type