24 May 2023 in IEEE Transactions on Intelligent Transportation Systems
Autonomous vehicle (AV) is expected to reshape the future transportation system, and its decision-making is one of the most critical modules. Many current decision-making modules are designed scenario-by-scenario and thus are not capable of meeting high-level AV requirements for lack of scalability to cope with the diversity of drivers’ demands and the infinity of traffic elements. This paper surveys the decision-making design inspired by driver intelligence and environment reasoning with better scalability for future high-level AVs. It involves three aspects: human factors in driving, environment reasoning, and detailed decision methods. The current state of the art of these three sections is surveyed in this paper. The characteristics of distinguished drivers, their decision mechanism, and the factors influencing decision-making are reviewed first to learn from excellent drivers. Environment reasoning is introduced following as a three-layer structure consisting of restriction, interaction, and attention. The optimization-based decision-making algorithms are then reviewed from the aspect of optimization targets, frameworks, applied scenarios, and limitations. Inspired by the existing research on driver intelligence and environment reasoning, a promising decision-making framework is also introduced for high-level AV design.
20 June 2023 in IEEE Transactions on Vehicular Technology
Autonomous Vehicles have wide-ranging applications in off-road environments. Off-road vehicular scenes can be abstracted as multi-agent systems, and trajectory prediction is a critical process for context understanding. Agents compete and cooperate other to pursue their individual targets, which makes trajectories complicated and changeable. Hence, in order to derive precise predictions, it is necessary to reason how agents interact with each other. However, current prediction algorithms lack a unified and appropriate method for spatiotemporal reasoning. Previous methods mainly rely on vehicle-lane constraints to capture features, which is only applicable in structured environments. To address this issue, this paper proposes a novel off-road multi-agent trajectory prediction framework called SA-LSTM. This framework comprises situation awareness extraction and Long Short Term Memory (LSTM) prediction backbones. Based on the analysis of agents' movement patterns, we decompose actions into the maintenance of previous movements and the strategy variations in response to environment situations. In situation awareness extraction, risk field and pooling layers are applied to filter interpretable awareness. As for prediction backbones, proper LSTM networks are selected to adapt the task features. In short-term predictions, a residual mechanism is used to preserve physical inertia, while for long-term predictions autoregression process is applied. We also establish a multi-agent dataset based on human-manipulated chase-and-run tasks to train and validate the performance of SA-LSTM. Experiments show that compared with the best baseline model at each prediction length, SA-LSTM reduces the mean absolute error by 7.27%, 32.99%, and 28.07% at the prediction time of 0.6, 3.0, and 6.0 seconds, respectively, proving better prediction accuracy.
7 August 2024 in IEEE Robotics and Automation Letters
This paper presents an efficient algorithm, naming Centralized Searching and Decentralized Optimization (CSDO), to find feasible solution for large-scale Multi-Vehicle Trajectory Planning (MVTP) problem. Due to the intractable growth of non-convex constraints with the number of agents, exploring various homotopy classes that imply different convex domains, is crucial for finding a feasible solution. However, existing methods struggle to explore various homotopy classes efficiently due to combining it with time-consuming, precise trajectory solution finding. CSDO, addresses this limitation by separating them into different levels and integrating an efficient Multi- Agent Path Finding (MAPF) algorithm to search homotopy classes. It first searches for a coarse initial guess using a large search step, identifying a specific homotopy class. Subsequent decentralized Sequential Quadratic Programming (SQP) refinement processes this guess, resolving minor collisions efficiently. Experimental results demonstrate that CSDO outperforms existing MVTP algorithms in large-scale, high-density scenarios, achieving up to a 95% success rate in 50 m × 50 m random scenarios around one second. Source codes are released in https://github.com/YangSVM/CSDOTrajectoryPlanning .
4 November 2024 in IEEE Transactions on Vehicular Technology
Planning module is an essential component of intelligent vehicle study. In this paper, we address the risk-aware planning problem of UGVs through a global-local planning framework which seamlessly integrates risk assessment methods. In particular, a global planning algorithm named Coarse2fine A* is proposed, which incorporates a potential field approach to enhance the safety of the planning results while ensuring the efficiency of the algorithm. A deterministic sampling method for local planning is leveraged and modified to suit off-road environment. It also integrates a risk assessment model to emphasize the avoidance of local risks. The performance of the algorithm is demonstrated through simulation experiments by comparing it with baseline algorithms, where the results of Coarse2fine A* are shown to be approximately 30% safer than those of the baseline algorithms, and local planning method demonstrates advantages in terms of safety and comfort compared to the baseline approachs. The practicality and effectiveness of the proposed planning framework are validated by deploying it on a real-world system consisting of a control center and a practical UGV platform.
9 September 2025 in Accident Analysis & Prevention
17 October 2025 in IEEE Transactions on Systems, Man, and Cybernetics: Systems
Scene understanding, defined as learning, extraction, and representation of interactions among traffic elements, is one of the critical challenges toward high-level autonomous driving (AD). Current scene understanding methods mainly focus on a single concrete task, such as trajectory prediction and risk level evaluation. Although they perform well on specific metrics, the generalization ability is insufficient to adapt to the complexity of real-world traffic and the diversity of downstream demands. In this study, we propose PreGSU, a generalized pretrained scene understanding model based on a graph attention network to learn the universal interaction and reasoning of traffic scenes to support various downstream tasks. After feature engineering and subgraph module processing, all elements are embedded as nodes to form a dynamic weighted graph. Then, four graph attention layers are applied to learn the relationships among agents and lanes. In the pretraining phase, the understanding model is trained on two self-supervised tasks: virtual interaction force (VIF) modeling and masked roadmap modeling (MRM). Based on the artificial potential field theory, VIF modeling enables PreGSU to capture the agent-to-agent interactions, while MRM extracts agent-to-road connections. In the fine-tuning process, the pretrained parameters are loaded to derive detailed understanding outputs. We conduct validation experiments on three datasets and two downstream tasks, i.e., trajectory prediction in urban scenario and intention recognition in highway scenario, to verify the model’s generalization and understanding capabilities. Results show that compared with single-task-driven baselines, PreGSU achieves competitive performance on all datasets and downstream tasks, indicating its potential to be generalized to various scenes and targets. Ablation study shows the effectiveness of pretraining task design.