United States Department of Transportation
i
Driver Behavior in Traffic
-
2012-12-01
-
[PDF-13.48 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Contracting Officer:
-
Corporate Publisher:
-
Abstract:Existing traffic analysis and management tools do not model the ability of drivers to recognize their environment and respond to it with behaviors that vary according to the encountered driving situation. The small body of literature on characterizing drivers behavior is typically limited to specific locations (i.e., by collecting data on specific intersections or freeway sections) and is very narrow in scope. This report documented the research performed to model driver behavior in traffic under naturalistic driving data. The research resulted in the development of hybrid car-following model. In addition, a neuro-fuzzy reinforcement learning, an agent-based artificial intelligence machine-learning technique, was used to model driving behavior. The naturalistic driving database was used to train and validate driver agents. The proposed methodology simulated events from different drivers and proved behavior heterogeneities. Robust agent activation techniques were also developed using discriminant analysis. The developed agents were implemented in VISSIM simulation platform and were evaluated by comparing the behavior of vehicles with and without agent activation. The results showed very close resemblance of the behavior of agents and driver data. Prototype agents prototype (spreadsheets and codes) were developed. Future research recommendations include training agents using more data to cover a wider region in the Wiedemann regime space, and sensitivity analysis of agent training parameters.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
