United States Department of Transportation
ROSA P serves as an archival repository of USDOT-published products including scientific findings, journal articles, guidelines, recommendations, or other information authored or co-authored by USDOT or funded partners.
As a repository, ROSA P retains documents in their original published format to ensure public access to scientific information.
i
Investigating the Effects of Cooperative Driving for CAVs in Different Driving Scenarios Using Multi-Driver Simulator Experiments
-
2022-08-29
[PDF-2.29 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Resource Type:
-
Geographical Coverage:
-
Corporate Publisher:
-
Abstract:Cooperative driving powered by connected vehicle (CV) technology is expected to improve traffic safety and efficiency, especially at locations with dense vehicle interactions. Although lots of research have developed their cooperative driving algorithms for different locations, the effects of human drivers in the loop and multi-agent driving decision making are less studied. In this project, three tasks were investigated: (1) developing cooperative driving strategies (CDS) for non-signalized intersections in a mixed traffic environment, and testing its effects in different automation level and market penetration rate; (2) proposing human-machine-interfaces (HMI) for non-signalized intersection cooperative driving, and evaluating the performance of different HMIs in various traffic conditions; (3) training a cooperative decision-making strategy for cooperative diverging at freeway off-ramp based on multi-agent reinforcement learning. UCF-SST self-developed human-in-the-loop co-simulation platform was used to complete the tasks. For task 1, an efficiency-oriented CDS was developed for mixed traffic cases, and tested on different CV and CAV market penetration rates. The experiment results showed that the proposed CDS reduced up to 53.8%, 66.4%, and 73.7% of travel time in CV-HDV (human-driven vehicle), CV-CAV, and CAV environments, respectively. For task 2, a driver-centered CDS was developed by modifying the algorithm in task 1, and then three different cooperative driving HMIs were evaluated by simulators. The results suggest a graphic-based HMI is better at displaying minor speed change requirements to the drivers, and it can guide the drivers approaching an intersection with better precision. For task 3, a multi-agent deep-Q network (MADQN) was trained for decision-making on freeway off-ramp diverging driving scenarios. The trained model significantly outperformed the baseline model in terms of efficiency and safety while ensuring decent successful diverging rate.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
