Скачать или смотреть Robot Learning, Navigation & Sensor Fusion - AI Frontiers 2025-10-17

This episode of AI Frontiers dives deep into the latest advancements in robotics research from October 17th, 2025. We explore how high-fidelity simulation, exemplified by GaussGym's use of 3D Gaussian Splatting, is revolutionizing robot training with unprecedented speed (over 100,000 steps/sec) and realism, significantly bridging the sim-to-real gap.

The research also tackles challenging autonomous navigation scenarios. Discover NOAH, a nature-inspired swarm optimization algorithm for underwater vehicles that accounts for currents and enables irreversible settlement, achieving an 86% success rate in anchoring. For disaster response, we examine an adaptive cost-map approach for path planning in partially unknown environments, allowing robots to intelligently interact with and push movable obstacles, improving goal-reach rates in cluttered spaces.

Robust sensor fusion is another key theme, with LVI-Q offering a tightly-coupled LiDAR-visual-inertial-kinematic odometry solution for quadruped robots, ensuring reliable localization even in difficult conditions. We also discuss ASBI, which leverages informative real-world data to actively tune black-box simulators, leading to more accurate models. Finally, PolyFly showcases advanced planning for collision-free cable-suspended aerial payload transportation, modeling components as polytopes for faster, non-conservative trajectories.

This synthesis was created using AI tools. The core content and summaries were generated by Google's Gemini 2.5 Flash Lite model, processing the provided paper summaries. Text-to-speech synthesis was performed using Deepgram, and accompanying imagery was generated by Grok. This AI-driven workflow allows for efficient and comprehensive coverage of cutting-edge research.

1. Jierui Peng et al. (2025). NEBULA: Do We Evaluate Vision-Language-Action Agents Correctly?. http://arxiv.org/pdf/2510.16263v2

2. Lukas Zbinden et al. (2025). Cosmos-Surg-dVRK: World Foundation Model-based Automated Online Evaluation of Surgical Robot Policy Learning. http://arxiv.org/pdf/2510.16240v1

3. Bihao Zhang et al. (2025). DeGrip: A Compact Cable-driven Robotic Gripper for Desktop Disassembly. http://arxiv.org/pdf/2510.16231v1

4. João Carlos Virgolino Soares et al. (2025). VAR-SLAM: Visual Adaptive and Robust SLAM for Dynamic Environments. http://arxiv.org/pdf/2510.16205v1

5. Zahra Arjmandi et al. (2025). Dynamic Recalibration in LiDAR SLAM: Integrating AI and Geometric Methods with Real-Time Feedback Using INAF Fusion. http://arxiv.org/pdf/2510.15803v1

6. Xinyue Xu et al. (2025). DexCanvas: Bridging Human Demonstrations and Robot Learning for Dexterous Manipulation. http://arxiv.org/pdf/2510.15786v1

7. Taehyeon Kim et al. (2025). Few-Shot Demonstration-Driven Task Coordination and Trajectory Execution for Multi-Robot Systems. http://arxiv.org/pdf/2510.15686v1

8. Yuhong Cao et al. (2025). HEADER: Hierarchical Robot Exploration via Attention-Based Deep Reinforcement Learning with Expert-Guided Reward. http://arxiv.org/pdf/2510.15679v1

9. Yameng Zhang et al. (2025). Freehand 3D Ultrasound Imaging: Sim-in-the-Loop Probe Pose Optimization via Visual Servoing. http://arxiv.org/pdf/2510.15668v1

10. Manuel J. Fernandez et al. (2025). Integration of a Variable Stiffness Link for Long-Reach Aerial Manipulation. http://arxiv.org/pdf/2510.15639v1

11. Jared K. Lepora et al. (2025). Educational SoftHand-A: Building an Anthropomorphic Hand with Soft Synergies using LEGO MINDSTORMS. http://arxiv.org/pdf/2510.15638v1

12. Hongyu Zhou et al. (2025). Adaptive Legged Locomotion via Online Learning for Model Predictive Control. http://arxiv.org/pdf/2510.15626v1

13. Shilei Li et al. (2025). Improved Extended Kalman Filter-Based Disturbance Observers for Exoskeletons. http://arxiv.org/pdf/2510.15533v1

14. Zehao Ni et al. (2025). VO-DP: Semantic-Geometric Adaptive Diffusion Policy for Vision-Only Robotic Manipulation. http://arxiv.org/pdf/2510.15530v2

15. Aron Distelzweig et al. (2025). Perfect Prediction or Plenty of Proposals? What Matters Most in Planning for Autonomous Driving. http://arxiv.org/pdf/2510.15505v1

16. Ziang Guo et al. (2025). VDRive: Leveraging Reinforced VLA and Diffusion Policy for End-to-end Autonomous Driving. http://arxiv.org/pdf/2510.15446v1

17. Zhaodong Yang et al. (2025). Towards Automated Chicken Deboning via Learning-based Dynamically-Adaptive 6-DoF Multi-Material Cutting. http://arxiv.org/pdf/2510.15376v1

18. Alejandro Escontrela et al. (2025). GaussGym: An open-source real-to-sim framework for learning locomotion from pixels. http://arxiv.org/pdf/2510.15352v1

19. Shyalan Ramesh et al. (2025). Nauplius Optimisation for Autonomous Hydrodynamics. http://arxiv.org/pdf/2510.15350v1

Disclaimer: This video uses arXiv.org content under its API Terms of Use; AI Frontiers is not affiliated with or endorsed by arXiv.org.