Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
102 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

A Survey on Autonomous Vehicle Control in the Era of Mixed-Autonomy: From Physics-Based to AI-Guided Driving Policy Learning (2007.05156v1)

Published 10 Jul 2020 in cs.AI and cs.RO

Abstract: This paper serves as an introduction and overview of the potentially useful models and methodologies from AI into the field of transportation engineering for autonomous vehicle (AV) control in the era of mixed autonomy. We will discuss state-of-the-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 AVdominated, and the pure AVs. This paper is primarily focused on the latter three phases. It is the first-of-its-kind survey paper to comprehensively review literature in both transportation engineering and AI for mixed traffic modeling. 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.

An Overview of Autonomous Vehicle Control in Mixed Autonomy

This paper provides a comprehensive survey of models and methodologies pertinent to autonomous vehicle (AV) control in mixed-autonomy scenarios, bridging transportation engineering and AI. It investigates the transition from traditional physics-based models to AI-driven policy learning, highlighting the unique challenges and opportunities these emerging technologies present. The paper categorizes AV deployment into four phases based on the dominance of automated versus human-driven vehicles (HVs) and focuses mainly on the latter three: HV-dominated, AV-dominated, and pure AVs.

Methodologies in AV Control

The survey underscores the importance of scalable driving policies, capable of coordinating AVs alongside human-driven vehicles within mixed traffic environments. It posits critical questions: the scalable management of AVs in mixed traffic, estimation of human driver behaviors, modeling uncontrolled AV behaviors, and characterizing interactions between AVs and HVs.

Game-Theoretic Approaches

The paper documents the application of game theory to model the strategic interactions between AVs and HVs. One-shot games are utilized for basic driving and lane-changes, while dynamic games extend this to continual motion planning, often employing Stackelberg models where AVs aim to predict HV responses. This approach emphasizes the significance of designing reward functions that account for both safety and efficiency, potentially steering HV behavior beneficially.

Reinforcement Learning Techniques

In conditions where AVs navigate alongside numerous HVs, the paper highlights reinforcement learning (RL) as a model-free strategy, offering adaptability in unpredictable environments. Both single-agent and multi-agent RL frameworks are discussed, reinforcing their utility in scenarios with partial observability and necessitating robust interactions between AVs and their environment.

Challenges in Mixed Autonomy

The paper identifies several challenges inherent in effective AV control within mixed traffic. These include the dynamic modeling of human behavior, the incorporation of heterogeneity among drivers, and the difficulty in navigating environments of unpredictable influences. It calls for a reevaluation of conventional traffic models, advocating for AI techniques that better capture human-like behaviors and decision-making processes.

Data Utilization and Model Validation

While discussing AV control, the paper underscores the crucial role data plays in shaping driving models. It provides insight into sources of trajectory data and suggests approaches for leveraging AI to refine these models, potentially via imitation learning algorithms like Generative Adversarial Imitation Learning (GAIL).

Moreover, the paper outlines the utility of simulators in model validation, emphasizing the need for platforms that facilitate interaction between AVs and HVs in controlled settings, allowing researchers to test hypotheses with high fidelity without real-world risks.

Future Directions and Implications

The integration of AI in AV control is poised to transform the landscape of transportation systems. The paper suggests further exploration into multi-scale modeling frameworks that accommodate complex, multi-agent dynamics and emergent traffic stability features. It calls for interdisciplinary collaboration to address open questions related to the ethics and societal impacts of AV deployment, ensuring these technologies enhance rather than compromise public safety.

Conclusion

This paper serves as a guidepost for researchers in the domain of AV control, advocating for AI-driven innovations to address the nuanced complexities of mixed traffic environments. By harmonizing vehicular interactions and leveraging AI's predictive capabilities, it anticipates significant advancements in traffic management efficiency and safety.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (2)
  1. Xuan Di (32 papers)
  2. Rongye Shi (12 papers)
Citations (164)