Modeling interactions in a collaborative connected automated transportation system: a game theory approach

Rahmati, Yalda

Permalink

https://hdl.handle.net/2142/115407 Copy

Description

Title

Modeling interactions in a collaborative connected automated transportation system: a game theory approach

Author(s)

Rahmati, Yalda

Issue Date

2022-04-21

Director of Research (if dissertation) or Advisor (if thesis)

Talebpour, Alireza

Doctoral Committee Chair(s)

Talebpour, Alireza

Committee Member(s)

• Ouyang, Yanfeng
• Meidani, Hadi
• Sowers, Richard B

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Connected Automated Vehicles
• Mixed Traffic
• Interactions
• Decision Modeling
• Human Behavior
• Game Theory
• Pedestrians
• Drivers

Abstract

Continuing evolution in automotive technology along with the growing number of related research in academia and industry gives the impression that self-driving cars are no longer a fantasy. Fueled by recent advances in computation capabilities, sensing, and navigation technologies, autonomous vehicles are envisaged to provide new levels of safety, mobility, and efficiency by taking human errors out of the driving equations. Yet, numerous critical hurdles remain, impeding the full (or even partial) operation of connected automated vehicles (CAVs) on public roads beyond the testing phases. Automated vehicles’ operation hinges on standard algorithms developed for robotic applications [1]. CAVs are, in essence, automated decision-making systems designed to perform driving tasks in a connected environment. Thanks to technology advancements, the problems of robot localization, control, and route finding in stationary environments around inanimate obstacles seem to be largely solved. However, the majority of robotic operations are designed and tested in confined environments with no/minimum human-robot interactions [2]. Proximity to humans introduces a new set of system complexities that justifies the need for a reliable technical framework to ensure safe and efficient human-robot coexistence. The advent of CAVs on urban roads is probably the most concrete example of robots operating in human spaces. CAVs promise a range of potential benefits, including, but not limited to, improving safety, efficiency, roadway capacity, and traffic flow stability in transportation networks. Realizing these objectives hinges on CAVs’ proper interaction with other road users as well as how the system, in general, reacts to their different driving behavior. In fact, in order to assess the overall impact of CAVs on traffic flow dynamics, human-CAV interactions in mixed traffic environments should be evaluated from both the humans’ and CAVs’ perspectives. From CAVs’ perspective, navigating in dynamic environments relies on estimating the future motion of surrounding obstacles and predicting potential conflicts [3]. However, the uncertainty, diversity, and subtlety of human behavior impose major challenges on predicting the future form of the environment in a transportation system. This becomes even more critical where no clear traffic rule defines priority, such as jaywalking pedestrians, parking lots, and unmarked intersections. Since direct human-CAV communication is not often possible, CAVs should resort to algorithms that use available sensory data to predict the future movement of surrounding road users and plan accordingly. A realistic model of human behavior is then vital to capture humans’ interactive behavior with others and provide the vehicle with accurate predictions of their future decisions. The second perspective in analyzing human-CAV interactions is to focus on human road users and explore how they might react to sharing roads with automated vehicles. A review of the literature indicates that one of the critical assumptions in mixed traffic research is consistency in human behavior over time. Most studies that have focused on CAVs’ long-term impact on transportation systems do not consider any changes in humans’ behavior under human-human and human-CAV interactions. An unanswered question here is whether human agents' behavior remains the same when interacting with automated vehicles on the road. Answering this question is particularly critical to fit more realistic models and better understand the future of transportation systems with the advent of automated vehicles. Based on the above discussion, this dissertation aims at developing the core concepts required to develop a collaborative, connected automated transportation system by (1) designing a reliable model of human behavior to use in CAVs’ controller system to enable the vehicle to predict the movement of surrounding human road users (e.g., pedestrians and drivers), and (2) investigating potential changes in humans’ behavior under human-human and human-CAV interactions and designing a technical framework to analyze the long-run equilibrium of the transportation system in the presence of automated vehicles. The first chapter of this dissertation defines the research problem and provides a detailed description of the research objectives that are expected to be accomplished. I then present a comprehensive review of the studies that have focused on human-CAV interactions in mixed traffic environments from both the humans' and CAVs' perspectives. Chapter 3 introduces the core of the modeling and analyses in this dissertation and delineates the general structure of the proposed modeling frameworks along with the mathematical representation of model calibration methods. Chapter 4 is dedicated to interactive decision making among pedestrians in real-world scenarios. The first step toward developing a general human behavior model for CAVs is to create a reliable model of pedestrian motion that can accurately capture the behavioral characteristics of pedestrians and predict their future movements given current conditions. In light of this, I have developed a novel game theory-based framework to predict humans’ walking decisions in a bidirectional pedestrian flow and assessed its performance in approximating crowd dynamics in microscopic and macroscopic levels. In the next chapter, I extended the aforementioned pedestrian motion model to also account for the effect of pedestrian-vehicle interactions on ones’ choice of walking strategies. I designed an interactive decision modeling framework that can be directly used by CAVs to predict the movement of surrounding pedestrians in shared environments and model their simultaneous decision-making when interacting with both vehicles and other pedestrians. After developing a reliable model of human behavior to predict pedestrians’ movements, a similar logic is adopted to create a modeling framework that enables CAVs to predict the interactive decision making among human drivers in complex driving scenarios. Such prediction capability is critical for realizing naturalistic autonomous driving and can potentially result in self-organized cooperative behaviors between drivers and CAVs, similar to those often observed between human-driven vehicles in complex driving environments. Chapters 7 and 8 analyze human-CAV interactions from humans' perspective and investigate consistency in humans’ driving behavior when interacting with automated vehicles. I have first focused on potential short-term changes in humans’ driving behavior with the advent of CAVs. To simplify the problem and have a more controlled testing environment, I have focused on the car-following behavior of human drivers in a mixed traffic environment. As discussed, analyzing potential shifts in drivers’ behavior under human-CAV interactions is a critical step in enhancing the realism of simulation frameworks and can facilitate planning for the future of transportation systems. Chapter 8 is dedicated to investigating the dynamics of human-CAV interactions over time and explores how CAVs' different driving characteristics might affect the driving norms of the future transportation system. The main contribution of this research is to introduce a general technical framework to identify the long-run equilibrium of traffic systems as humans start sharing roads with automated vehicles. The proposed framework is the first step toward assessing the long-term impacts of the CAV technology on safety, congestion, traffic stream stability, and energy consumption. Finally, the dissertation is concluded with summary remarks and future research directions.

Graduation Semester

2022-05

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Yalda Rahmati

Owning Collections