Autonomous Driving at Unsignalized Intersections: A Review of Decision-Making Challenges and Reinforcement Learning-Based Solutions

TL;DR

The paper addresses the challenge of autonomous driving at unsignalized intersections, a domain with high multi-agent uncertainty and safety-critical implications. It surveys reinforcement learning and deep learning–based decision-making approaches, analyzes driver intention inference, and frames the problem with MDPs and POMDPs to account for partial observability. Key contributions include a structured literature review, identification of limitations such as simulation-only validation and lack of end-to-end feasibility with motion planning, and proposed directions for robust, real-world applicable architectures. The work emphasizes hierarchical decision-making that couples high-level behavioral planning with low-level motion control, highlights sim-to-real transfer methods, and calls for real-world experimentation with high-fidelity vehicle dynamics to advance practical deployment. Overall, the survey provides a roadmap for developing safe, feasible, and generalizable decision-making systems for unsignalized intersection traversal.

Abstract

Autonomous driving at unsignalized intersections is still considered a challenging application for machine learning due to the complications associated with handling complex multi-agent scenarios characterized by a high degree of uncertainty. Automating the decision-making process at these safety-critical environments involves comprehending multiple levels of abstractions associated with learning robust driving behaviors to enable the vehicle to navigate efficiently. In this survey, we aim at exploring the state-of-the-art techniques implemented for decision-making applications, with a focus on algorithms that combine Reinforcement Learning (RL) and deep learning for learning traversing policies at unsignalized intersections. The reviewed schemes vary in the proposed driving scenario, in the assumptions made for the used intersection model, in the tackled challenges, and in the learning algorithms that are used. We have presented comparisons for these techniques to highlight their limitations and strengths. Based on our in-depth investigation, it can be discerned that a robust decision-making scheme for navigating real-world unsignalized intersection has yet to be developed. Along with our analysis and discussion, we recommend potential research directions encouraging the interested players to tackle the highlighted challenges. By adhering to our recommendations, decision-making architectures that are both non-overcautious and safe, yet feasible, can be trained and validated in real-world unsignalized intersections environments.

Figures (11)

• Figure 1: Different types of unsignalized intersections. These images are generated using SUMO (Simulation of Urban MObility) traffic simulation software.
• Figure 2: Surveyed decision-making challenges and solutions.
• Figure 3: Decision-making processes in urban autonomous vehicles.
• Figure 4: An intersection-traversal scenario where the ego vehicle is required to handle several sorts of uncertainties associated with the approaching vehicle.
• Figure 5: LSTM for solving the formulated POMDP of intersection-traversal problem.