A Survey on Autonomous Vehicle Control in the Era of Mixed-Autonomy: From Physics-Based to AI-Guided Driving Policy Learning

Abstract

This paper serves as an introduction and overview of the potentially useful models and methodologies from artificial intelligence (AI) into the field of transportation engineering for autonomous vehicle (AV) control in the era of mixed autonomy. It is the first-of-its-kind survey paper to comprehensively review literature in both transportation engineering and AI for mixed traffic modeling. We will discuss state-ofthe-art applications of AI-guided methods, identify opportunities and obstacles, raise open questions, and help suggest the building blocks and areas where AI could play a role in mixed autonomy. We divide the stage of autonomous vehicle (AV) deployment into four phases: the pure HVs, the HV-dominated, the AV-dominated, and the pure AVs. This paper is primarily focused on the latter three phases. Models used for each phase are summarized, encompassing game theory, deep (reinforcement) learning, and imitation learning. While reviewing the methodologies, we primarily focus on the following research questions: (1) What scalable driving policies are to control a large number of AVs in mixed traffic comprised of human drivers and uncontrollable AVs? (2) How do we estimate human driver behaviors? (3) How should the driving behavior of uncontrollable AVs be modeled in the environment? (4) How are the interactions between human drivers and autonomous vehicles characterized? Hopefully this paper will not only inspire our transportation community to rethink the conventional models that are developed in the data-shortage era, but also reach out to other disciplines, in particular robotics and machine learning, to join forces towards creating a safe and efficient mixed traffic ecosystem.

Leading Cruise Control in Mixed Traffic Flow: System Modeling, Controllability, and String Stability

“…Remark 1: In the general LCC system (9), the CF-LCC system (11) and the CCC system (13), the velocity error of the head vehicle ṽh (t) can be viewed as an external disturbance signal into the system. Note that the models ( 9), ( 11), (12), and ( 13) are all based on an linearization around an equilibrium velocity v * that needs to be chosen or designed carefully. For LCC, CF-LCC and CCC, the equilibrium velocity v * is determined by the equilibrium velocity of the head vehicle v h , and this information can be estimated from its historical trajectory data or current traffic condition.…”

“…Theorem 1: The FD-LCC system with no vehicle ahead and n HDVs behind, given by (12), is controllable, if the following condition holds…”

“…We utilize these two metrics to measure the energy-related control difficulty of the LCC systems at different system sizes. We focus on the FD-LCC system (12), which can be abstracted as a chain network with one single input node and n uncontrolled nodes, as shown in Fig. 4.…”

“…This signal might be easier to obtain compared to the spacing error signal si (t), since the equilibrium spacing s * of one HDV is nontrivial to measure accurately. We now consider the FD-LCC system (12), and the output is given by…”

“…Theorem 3: Consider the FD-LCC system with no vehicle ahead and n HDVs behind, given by (12). Suppose the CAV can measure the velocity error of vehicle k (1 ≤ k < n), given by (24).…”