Abstract

Autonomous Vehicles (AVs) are expected to transform transportation in the near future. Although considerable progress has been made, widespread adoption of AVs will not become a reality until solutions are developed that enable AVs to co-exist with Human-driven Vehicles (HVs). There are still many challenges preventing Connected and Autonomous Vehicles (CAVs) from safely and smoothly navigating. We identify two major challenges in this direction. First, the communication system is not always reliable and suffers from noise and information loss. Second, AV navigation in the presence of HVs is challenging, as HVs continuously update their policies in response to AVs and the social preferences and behaviors of human drivers are unknown. Towards this end, we first propose solutions to improve situational awareness by enabling reliable and robust Cooperative Vehicle Safety (CVS) systems that mitigate the effect of information loss and propose a hybrid learning-based predictive modeling technique for CVS systems. Our prediction system is based on a Hybrid Gaussian Process (HGP) approach that provides accurate vehicle trajectory predictions to compensate for information loss. We use offline real-world data to learn a finite bank of driver models that represent the joint dynamics of the vehicle and the driver's behavior. AVs and HVs equipped with such reliable vehicular communication can coordinate, improving safety and efficiency. However, even in the presence of perfect communication, is still challenging for CAVs to navigate in the presence of humans. Therefore, we study the cooperative maneuver planning problem in a mixed autonomy environment. We frame the mixed-autonomy problem as a Multi-Agent Reinforcement Learning (MARL) problem and propose an approach that allows AVs to learn the decision-making of HVs implicitly from experience, account for all vehicles' interests, and safely adapt to other traffic situations. In contrast with existing works, we quantify AVs' social preferences and propose a distributed reward structure that introduces altruism into their decision-making process, allowing the altruistic AVs to learn to establish coalitions and influence the behavior of HVs. Inspired by humans, we provide our AVs with the capability of anticipating future states and leveraging prediction in the MARL decision-making framework. We propose the integration of two essential components of AVs, i.e, social navigation and prediction, and present a prediction-aware planning and social-aware optimization RL framework. Our proposed framework take advantage of a Hybrid Predictive Network (HPN) that anticipates future observations. The HPN is used in a multi-step prediction chain to compute a window of predicted future observations to be used by the Value Function Network (VFN). Finally, a safe VFN is trained to optimize a social utility using a sequence of previous and predicted observations, and a safety prioritizer is used to leverage the predictions to mask the unsafe actions, constraining the RL policy. The experiments on real-world and simulated data demonstrated the performance improvement of the proposed solutions in both safety and traffic-level metrics and validate the advantages and applicability of our solutions.

Notes

If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu

Graduation Date

2022

Semester

Fall

Advisor

Pourmohammadi Fallah, Yaser

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Electrical and Computer Engineering

Degree Program

Computer Engineering

Identifier

CFE0009842; DP0027783

URL

https://purls.library.ucf.edu/go/DP0027783

Language

English

Release Date

June 2028

Length of Campus-only Access

5 years

Access Status

Doctoral Dissertation (Campus-only Access)

Download

Restricted to the UCF community until June 2028; it will then be open access.

Share

COinS