The field of autonomous driving and robotics is witnessing significant advancements with a focus on improving safety, efficiency, and performance. Researchers are exploring hybrid approaches that combine the strengths of rule-based and learning-based planners to achieve better generalization and real-time performance. Another notable trend is the development of retrieval-augmented generation frameworks that enable fine-grained control and safe behavior across diverse scenarios. Confidence-guided human-AI collaboration is also gaining traction, allowing for rapid and stable learning of human-guided policies with minimal human interaction. Furthermore, enhancing the safety of foundation models for visual navigation through collision avoidance techniques is becoming increasingly important. Noteworthy papers include:

• SAH-Drive, which proposes a scenario-aware hybrid planner for closed-loop vehicle trajectory generation, achieving state-of-the-art performance while maintaining computational efficiency.
• RealDrive, which introduces a retrieval-augmented generation framework that initializes a diffusion-based planning policy by retrieving the most relevant expert demonstrations from the training dataset, resulting in improved generalization to long-tail events and enhanced trajectory diversity.
• Confidence-Guided Human-AI Collaboration, which develops a strategy to overcome safe-exploration and distribution-shift challenges in autonomous driving, achieving state-of-the-art results in terms of safety, efficiency, and overall performance.
• CARE, which proposes a plug-and-play module that enhances the safety of vision-based navigation without requiring additional range sensors or fine-tuning of pretrained models, consistently reducing collision rates without sacrificing goal-reaching performance.
• LiPo, which presents a lightweight post-optimization framework for smoothing chunked action sequences, significantly reducing vibration and motion jitter, leading to smoother execution and improved mechanical robustness.