Abstract:The decision-making algorithm serves as a fundamental component for advancing the level of autonomous driving. The end-to-end decision-making algorithm has a strong ability to process the original data, but it has grave uncertainty. However, other learning-based decision-making algorithms rely heavily on ideal state information and are entirely unsuitable for autonomous driving tasks in real-world scenarios with incomplete global information. Addressing this research gap, this paper proposes a stable hierarchical decision-making framework with images as the input. The first step of the framework is a model-based data encoder that converts the input image data into a fixed universal data format. Next is a state machine based on a time series Graph Convolutional Network (GCN), which is used to classify the current driving state. Finally, according to the state's classification, the corresponding rule-based algorithm is selected for action generation. Through verification, the algorithm demonstrates the ability to perform autonomous driving tasks in different traffic scenarios without relying on global network information. Comparative experiments further confirm the effectiveness of the hierarchical framework, model-based image data encoder, and time series GCN.

What problem does this paper attempt to address?

This paper attempts to address the problem of implementing decision-making algorithms for autonomous driving in partially connected traffic scenarios. Specifically, existing end-to-end decision-making algorithms, while capable of handling raw data, exhibit significant uncertainty; other learning-based decision-making algorithms heavily rely on ideal state information, making them unsuitable for complex traffic environments with incomplete information in the real world. Therefore, this paper proposes a stable hierarchical decision-making framework based on image input, aiming to overcome these limitations and achieve autonomous driving tasks in various traffic scenarios. ### Main Issues: 1. **Uncertainty of end-to-end decision-making algorithms**: Although end-to-end decision-making algorithms can directly generate vehicle actions from raw data, their outputs are highly uncertain, making it difficult to ensure safety and stability in practical applications. 2. **Dependence of existing learning-based decision-making algorithms on ideal information**: Most learning-based decision-making algorithms require ideal state information, which is difficult to obtain in the real world, especially in traffic scenarios with incomplete information. 3. **Insufficient ability to adapt to incomplete information**: Existing algorithms perform poorly in handling complex and dynamic traffic scenarios, particularly when information is incomplete. ### Solutions: 1. **Hierarchical decision-making framework**: This paper proposes a hierarchical decision-making framework that converts image data into a fixed-format state matrix, uses a temporal graph convolutional network (GCN) for driving state classification, and finally selects the corresponding rule-based algorithm to generate actions based on the classification results. 2. **Data encoder based on pre-trained models**: A pre-trained object detection model is used to preprocess image data, converting it into a state matrix suitable for the decision-making network, and adding directional coordinates to enhance the algorithm's adaptability to perceptual information. 3. **Temporal GCN**: A temporal graph convolutional network is introduced to effectively capture temporal dependencies, improving the algorithm's understanding of the scene. 4. **Observable and controllable process**: Compared to traditional end-to-end models, this framework makes the entire process of perceptual data preprocessing, driving state classification, and action generation more transparent and controllable, enhancing stability and interpretability. Through these innovations, this paper aims to improve the adaptability and robustness of autonomous driving systems in partially connected traffic scenarios, thereby better addressing the complex traffic environments of the real world.

Learning-Based Hierarchical Decision-Making Framework for Automatic Driving in Incompletely Connected Traffic Scenarios

A Hierarchical Imitation Learning-based Decision Framework for Autonomous Driving

Decision making framework for autonomous vehicles driving behavior in complex scenarios via hierarchical state machine

A hierarchical perception decision-making framework for autonomous driving

Convolutional Neural Network-Based Intelligent Decision-Making for Automated Vehicles

Spatio-Temporal Image Representation and Deep-Learning-Based Decision Framework for Automated Vehicles

HKTSG: A Hierarchical Knowledge-guided Traffic Scene Graph Representation Learning Framework for Intelligent Vehicles

Deep Reinforcement Learning Based Decision-Making Strategy of Autonomous Vehicle in Highway Uncertain Driving Environments

Vision-Based Autonomous Driving: A Hierarchical Reinforcement Learning Approach

Decision-Making Framework for Automated Driving in Highway Environments

Learning a Decision Module by Imitating Driver's Control Behaviors

Humanlike Driving: Empirical Decision-Making System for Autonomous Vehicles

Hierarchical reinforcement learning for self-driving decision-making without reliance on labelled driving data

Deep Reinforcement Learning Based High-level Driving Behavior Decision-making Model in Heterogeneous Traffic

Integrated Decision and Control: Towards Interpretable and Computationally Efficient Driving Intelligence

Integrated Decision and Control: Toward Interpretable and Computationally Efficient Driving Intelligence

Deep Reinforcement Learning Enabled Decision-Making for Autonomous Driving at Intersections

Behavior Decision-making Method for Autonomous Vehicle

Adaptive Decision Making at the Intersection for Autonomous Vehicles Based on Skill Discovery

Deep Q-Network Based Decision Making for Autonomous Driving

Decision-making Strategy on Highway for Autonomous Vehicles using Deep Reinforcement Learning