Dachuan Li

Tactical decision-making plays a key role in ensuring the operational safety of autonomous vehicle (AV) systems. Recent environment perception techniques are built upon the deep learning (DL) approaches and utilize probabilistic representations to quantitatively capture the inherent uncertainties in networks. The downstream decision-making module shall utilize the perceived probabilistic information to derive proper maneuver decisions. However, how to translate the uncertainties of DL between the two stages to manage the risk remains an open problem. This paper proposes a holistic and interpretable framework that can explicitly represent uncertainties of DL-based perception and generate safe maneuvers by accounting for such uncertainties. By quantifying and propagating uncertainties of perceptions networks, an explicit link is established between DL perception uncertainty and safety utility of decision-making via the Adaptive Likelihood mechanism. Built on this basis, we propose an uncertainty-aware decision-making approach (PUALD-Net) that can incorporates both regression and classification perception uncertainties into probabilistic reasoning and derive risk-bounding decisions. Real-world field test results demonstrate that the proposed framework can effectively ensure safety, and outperforms existing approaches in terms of safety and driving capability metrics (Infraction Rate and Driving Score) in adverse scenarios incurring high perception uncertainties.

Reinforcement learning (RL) offers a powerful paradigm for adaptive motion planning in complex driving environments. However, applying RL to autonomous driving remains challenging due to uncertainty from partial observability and the stochastic, multimodal behaviors of traffic participants. This paper presents a novel risk-aware RL framework for non-conservative motion planning under uncertainty. By integrating Partially Observable Markov Decision Processes (POMDP) with a deep RL-based policy optimization scheme, the proposed approach explicitly models aleatoric uncertainty via a Gaussian Mixture Bayesian Belief Updater and a time-varying risk field. Additionally, an Adaptive Context-aware Attention (ACA) module is employed to prioritize critical targets for enhanced interaction modeling dynamically. Extensive experiments on the CARLA simulator show that the framework generalizes well across diverse traffic conditions, improving average reward by 65.74% and 64.02% in low-speed dense and high-speed sparse scenarios. It remains robust in challenging situations such as overtaking and sudden lane changes in the PeMS dataset. Furthermore, distributed deployment tests confirm a real-time performance of 10 Hz on a hardware-in-the-loop platform, demonstrating the feasibility of practical deployment.

Automated Vehicles (AVs) will coexist with Human-Driven Vehicles (HDVs) for a long time. AVs must navigate safely among HDVs while maintaining smooth traffic flow. To facilitate this, the decision making system of AVs must accurately assess HDV intentions while accounting for inherent uncertainties. Current HDV intention prediction models often misclassify these intentions, leading to unsafe navigation decisions. This study introduces a three-stage Bayesian sequential game-based decision making architecture designed for AV operation. In the first stage, the AV utilizes a temporal neural network to classify vehicle intentions. In the second stage, a sequential game is solved to determine optimal actions by predicting future HDV states. The final stage, serving as a validation stage, identifies and corrects misclassifications from the first stage by predicting HDV future positions, incorporating models that account for potential deviations from the ground truth. Simulation results indicate a 93.5±0.5% accuracy in initial intention predictions, facilitating swift and effective decision making. The validation stage further enhances safety by promptly correcting errors, ensuring reliable navigation for AVs in HDV environments.

Effectively summarizing dense 3D point cloud data and extracting motion information of moving objects (moving object segmentation, MOS) is crucial to autonomous driving and robotics applications. How to effectively utilize motion and semantic features and avoid information loss during 3D-to-2D projection is still a key challenge. In this paper, we propose a novel multi-view MOS model (MV-MOS) by fusing motion-semantic features from different 2D representations of point clouds. To effectively exploit complementary information, the motion branches of the proposed model combines motion features from both bird's eye view (BEV) and range view (RV) representations. In addition, a semantic branch is introduced to provide supplementary semantic features of moving objects. Finally, a Mamba module is utilized to fuse the semantic features with motion features and provide effective guidance for the motion branches. We validated the effectiveness of the proposed multi-branch fusion MOS framework via comprehensive experiments, and our proposed model outperforms existing state-of-the-art models on the SemanticKITTI benchmark.

The accurate prediction of behaviors of surrounding traffic participants is critical for autonomous vehicles (AV). How to fully encode both explicit (e.g., map structure and road geometry) and implicit scene context information (e.g., traffic rules) within complex scenarios is still challenging. In this work, we propose an implicit scene context-aware trajectory prediction framework (the PRISC-Net, Prediction with Implicit Scene Context) for accurate and interactive behavior forecasting. The novelty of the proposed approach includes: 1) development of a behavior prediction framework that takes advantage of both model- and learning-based approaches to fully encode scene context information while modeling complex interactions; 2) development of a candidate path target predictor that utilizes explicit and implicit scene context information for candidate path target prediction, along with a motion planning-based generator that generates kinematic feasible candidate trajectories; 3) integration of the proposed target predictor and trajectory generator with a learning-based evaluator to capture complex agent-agent and agent-scene interactions and output accurate predictions. Experiment results based on vehicle behavior datasets and real-world road tests show that the proposed approaches outperform state-of-the-art methods in terms of prediction accuracy and scene context compliance.

This paper advances state-of-the-art automated driving systems with a comprehensive framework that encompasses decision making, maneuver planning, and trajectory tracking considering safety, computational efficiency, and passenger comfort. In face of the co-existence of automated vehicles (AVs) and human-driven vehicles (HDVs), a decision making framework of AVs is proposed for safe lane keeping or changing. The decision making is based on the HDVs' future motion predicted by a learning-based Long Short-Term Memory model. To quantify the uncertainties in prediction, an error ellipse is used to capture the model deviations from the ground truth to ensure driving safety. This paper develops a novel method that leverages lower-order parametric curves to efficiently generate feasible, safe, and comfortable lateral movements for AVs. The planner is complemented by maneuver replanning that can guide the AV back to the original lane when confronted with unexpected blockages from surrounding vehicles. Based on real-world datasets, simulation results show that the proposed method achieves curvature compatibility, shorter trajectory length in lateral maneuvers, accurate trajectory tracking, and effective collision avoidance in lane changing.

Conventional perception-planning pipelines of autonomous vehicles (AV) utilize deep learning (DL) techniques that typically generate deterministic outputs without explicitly evaluating their uncertainties and trustworthiness. Therefore, the downstream decision-making components may generate unsafe outputs leading to system failure or accidents, if the preceding perception component provides highly uncertain information. To mitigate this issue, this paper proposes a coherent safe perception-planning framework that quantifies and transfers DL-based perception uncertainties. Following the Bayesian Deep Learning paradigm, we design a probabilistic 3D object detector that extracts objects from LiDAR point clouds while quantifying the corresponding aleatoric and epistemic uncertainty. A chance-constrained motion planner is designed to formulate an explicit link between DL-based perception uncertainties and operation risk and generate safe and risk-bounding trajectories. The proposed framework is validated through various challenging scenarios in the CARLA simulator. Experiment results demonstrate that our framework can effectively capture the uncertainties in DL, and generate trajectories that bound the risk under DL perception uncertainties. It also outperforms counterpart approaches without explicitly evaluating the uncertainties of DL-based perception.