Learning to Drive at Unsignalized Intersections using Attention-based Deep Reinforcement Learning

Hyunki, Seong,; Jung, Choong–Hwan; Lee, Seung-Wook; Shim, David Hyunchul

doi:10.1109/itsc48978.2021.9564720

Cited by 25 publications

(8 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Gonzalez et al 22 reported a human-like decision approach using Partially Observable Markov Decision Process (POMDP), which can mimic the human abilities of anticipating surrounding drivers' intentions in highway driving. Seong et al 23 proposed an attention-based deep reinforcement learning framework for interactive driving at unsignalized intersections, which can realize human-likeness by learning to focus on spatially and temporally important features. Another way to realize human-likeness is to incorporate human decision rules, styles, or preferences in mechanism-based algorithms.…”

Section: Mechanism-basedmentioning

confidence: 99%

Human-like driving decision at unsignalized intersections based on game theory

Liu

Xiao

2022

Proceedings of the Institution of Mechanical Engineers, Part D:

View full text Add to dashboard Cite

Unsignalized intersection driving is challenging for automated vehicles. For safe and efficient performances, the diverse and dynamic behaviors of interacting vehicles should be considered. Based on a game-theoretic framework, a human-like payoff design methodology is proposed for the automated decision at unsignalized intersections. Prospect Theory is introduced to map the objective collision risk to the subjective driver payoffs, and the driving style can be quantified as a tradeoff between safety and speed. To account for the dynamics of interaction, a probabilistic model is further introduced to describe the acceleration tendency of drivers. Simulation results show that the proposed decision algorithm can describe the dynamic process of two-vehicle interaction in limit cases. Statistics of uniformly-sampled cases simulation indicate that the success rate of safe interaction reaches 98%, while the speed efficiency can also be guaranteed. The proposed approach is further applied and validated in four-vehicle interaction scenarios at a four-arm intersection.

show abstract

Section: Mechanism-basedmentioning

confidence: 99%

Human-like driving decision at unsignalized intersections based on game theory

Liu

Xiao

2022

Proceedings of the Institution of Mechanical Engineers, Part D:

View full text Add to dashboard Cite

show abstract

“…In recent years, considerable research has been devoted to the problems of decision-making and interaction of AV in intersection scenarios. These studies have employed various approaches, including ruled-based methods [12], gametheoretic methods [5], [13], and data-driven techniques [8], [9], [14], in which RL is recognized as a flexible, efficient, and potent method. However, its widespread implementation is hindered by several obstacles, one of which is training a RL model that can effectively manage a range of driving situations and decision-making tasks [8].…”

Section: A Decision-making Of Av At Intersectionmentioning

confidence: 99%

“…Reinforcement learning (RL), due to its exceptional learning capabilities and computational efficiency, has been widely applied in the design of decision-making algorithms for AV [7]- [9]. However, these methods mostly use a single model to handle different tasks, and utilizing a single model to cope with multiple autonomous driving decision-making scenarios and tasks remains a significant challenge for RL [8].…”

Section: Introductionmentioning

confidence: 99%

State Key Laboratory for Geomechanics and Deep Underground Engineering, School of Mechanics & Civil Engineering, China University of Mining & Technology, Xuzhou 221116, China

Liu

Wei

Wang

et al. 2019

Mathematical Biosciences and Engineering

View full text Add to dashboard Cite

This paper is devoted to the L p (p > 1) solutions of one-dimensional backward stochastic differential equations (BSDEs for short) with general time intervals and generators satisfying some non-uniform conditions in t and ω. An existence and uniqueness result, a comparison theorem and an existence result for the minimal solutions are respectively obtained, which considerably improve some known works.Some classical techniques used to deal with the existence and uniqueness of L p (p > 1) solutions of BSDEs with Lipschitz or linear-growth generators are also developed in this paper.

show abstract

“…To mitigate the threat of potential vehicle failures to traffic safety and efficiency, Pei et al constructed a rule-based fault-tolerant cooperative driving strategy for signal-free intersections by modeling potential vehicle failure types to balance traffic safety and efficiency [2]. However, rule-based methods often employ conservative control schemes to avoid conflicts [3], which may not effectively maximize the utilization of space resources at intersection. Mathematical optimization techniques typically establish mathematical models based on specific optimization objectives and obtain optimal traffic control strategies by solving the optimal solution.…”

Section: Introductionmentioning

confidence: 99%

“…This alleviates computational load, leading to a more efficient training process. (3) Simulation experiments are conducted under low, medium, and high traffic flow conditions to validate the efficiency, safety, and driving comfort of the proposed algorithm.…”

Section: Introductionmentioning

confidence: 99%

Autonomous vehicle scheduling at signal-free intersections based on deep reinforcement learning

He,

Hao,

Guan

2024

International Conference on Smart Transportation and City Engineering (STCE 2023)

View full text Add to dashboard Cite

Coordinating autonomous vehicles (AVs) at signal-free intersections has emerged as a critical area of research in intelligent transportation. This paper presents a novel vehicle scheduling strategy based on multi-agent deep reinforcement learning (MADRL) to enhance traffic efficiency and reduce collision rates at signal-free intersections. The strategy incorporates the use of virtual lane techniques to simplify the management of vehicle trajectory conflicts. Furthermore, vehicle platooning techniques are employed to alleviate the computational burden on the reinforcement learning controller by treating multiple AVs as a single agent, thereby optimizing training performance. The MADRL approach is adopted to facilitate platoon coordination and enhance intersection traffic efficiency. Through comprehensive simulations, the proposed approach demonstrates its effectiveness in improving traffic efficiency, driving comfort, and safety at signal-free intersections.

show abstract

Learning to Drive at Unsignalized Intersections using Attention-based Deep Reinforcement Learning

Cited by 25 publications

References 15 publications

Human-like driving decision at unsignalized intersections based on game theory

Human-like driving decision at unsignalized intersections based on game theory

State Key Laboratory for Geomechanics and Deep Underground Engineering, School of Mechanics & Civil Engineering, China University of Mining & Technology, Xuzhou 221116, China

Autonomous vehicle scheduling at signal-free intersections based on deep reinforcement learning

Contact Info

Product

Resources

About