Recent advances in Large Vision–Language Models (LVLMs) have demonstrated strong cross-modal reasoning capabilities, offering new opportunities for decision-making in autonomous driving. However, existing end-to-end approaches still suffer from limited semantic consistency, weak task controllability, and insufficient interpretability. To address these challenges, we propose SemAlign-E2E (Semantic-Aligned End-to-End), a semantic-aligned multimodal LVLM framework that unifies visual, LiDAR, and task-oriented textual inputs through cross-modal attention. This design enables end-to-end reasoning from scene understanding to high-level driving command generation. Beyond producing structured control instructions, the framework also provides natural-language explanations to enhance interpretability. We conduct extensive evaluations on the nuScenes dataset and CARLA simulation platform. Experimental results show that SemAlign-E2E achieves substantial improvements in driving stability, safety, multi-task generalization, and semantic comprehension, consistently outperforming state-of-the-art baselines. Notably, the framework exhibits superior behavioral consistency and risk-aware decision-making in complex traffic scenarios. These findings highlight the potential of LVLM-driven semantic reasoning for autonomous driving and provide a scalable pathway toward future semantic-enhanced end-to-end driving systems.

The long-term coexistence of human-driven vehicles (HVs) and autonomous vehicles (AVs) in mixed traffic presents significant challenges for lane-change interactions on freeways. To address this, we propose a closed-loop decision-making framework, centered on Social Value Orientation (SVO), that covers the entire process from recognition to fallback execution. First, we use maximum-entropy inverse reinforcement learning (MaxEnt-IRL) to infer driver SVO parameters (${\theta }_{SVO}$) from the NGSIM dataset, quantifying the trade-off between selfish and cooperative behaviors as learnable weights. These parameters are then incorporated into a Transformer-based predictor via conditional embeddings, enabling the model to generate personalized trajectories from identical historical data. Furthermore, within a receding-horizon, game-theoretic framework, we combine preference-weighted payoffs with this conditional predictor and introduce a dynamic lane-change abort mechanism. This mechanism triggers a fallback maneuver, generated by an APF + MPC controller, if the expected return of continuing the lane change drops below that of aborting. Simulations across 1000 adversarial scenarios show that our method markedly improves the lane-change success rate and cruising efficiency compared to the IDM + MOBIL baseline. It also significantly reduces forced merges and hazardous events when encountering aggressive or selfish blocking vehicles, demonstrating the safety and robustness benefits of our preference-aware model and abort mechanism.

To address the challenge of directly translating predicted distributions of dynamic obstacles into three-dimensional (3D) planning constraints during unmanned aerial vehicle (UAV) flight, we propose Torch GIP-RB-SI-3DA*, a graph-interaction-prediction-based method for planning safe flight intervals and dynamic 3D paths. The model predicts trajectory distributions over a 15 s horizon from the 3D positions and velocities observed in six consecutive frames and maps the calibrated distributions to instantaneous voxel-level risks. The front end enforces both instantaneous- and cumulative-risk constraints through multi-label search. The back end performs finite-difference smoothing and joint retiming, followed by dense-sampling verification of collision freedom and the risk, velocity, acceleration, and jerk constraints. Across 10 independent training seeds, Torch GIP reduces the average displacement error (ADE) and final displacement error (FDE) by 5.07–15.25% and 5.93–9.29%, respectively, relative to NoGraph, while also achieving lower negative log-likelihood (NLL). All these improvements remain statistically significant after Holm correction. In 100 pre-screened in-distribution (ID) scenarios, Torch GIP generates 88 candidate trajectories, of which 82 are collision-free under ground truth, 76 satisfy the ground-truth instantaneous-risk constraint, and 82 satisfy the ground-truth risk budget in end-to-end evaluation. Although its candidate yield is lower than that of the constant-velocity (CV) baseline, its yields under all three ground-truth evaluation criteria are higher than those of both CV and rule-based GIP. Paired ablation studies identify joint retiming as a critical component for maintaining end-to-end yield. Across 50 out-of-distribution (OOD) scenarios, the method generates 48 planned trajectories, of which 41 are collision-free under ground truth, and 33 satisfy the ground-truth risk budget. All 16 replays of UZH-FPV trajectories with known future motion pass both offline and sampling-based verification. These results demonstrate that the proposed method provides a systematic and verifiable transformation from predicted distributions to 3D planning constraints.

At autonomous intersections in mixed traffic, where Connected and Autonomous Vehicles (CAVs) coexist with Human-Driven Vehicles (HDVs), scheduling must remain effective even when a reliable explicit HDV turning intention is not available sufficiently early for the scheduling decision. This paper proposes a learning-based platoon scheduling method for this information-limited setting. CAVs and HDVs are organized into mixed platoons or an HDV group, and a state-augmentation function is designed to encode their temporal relationship while preserving the priority of uncontrollable HDVs. An enumeration (EN)-based expert demonstration mechanism is further integrated into the training process to provide high-quality experience. In the representative training run reported in the manuscript, the expert-assisted model achieved a peak average reward approximately 14% higher than the model trained without demonstrations. Across the tested demand cases, the learned scheduler also obtained travel-cost performance close to the EN benchmark. The scope of the method is explicitly limited to the modeled lane-keeping and sensing assumptions; robustness to random seeds, unexpected HDV maneuvers, communication delays, and additional safety metrics requires dedicated validation.