analiza aiqbench
EmbodiedDiffusion advances autonomous navigation by unifying traversability estimation and trajectory generation within a single diffusion model that eliminates the need for separate planners or heavy mapping. By distilling semantics from Vision-Language Models into lightweight student networks, it enables real-time inference on heterogeneous platforms like quadrupeds and aerial robots with rapid adaptation to new environments. This approach represents a significant step toward scalable, prompt-free robotic systems capable of cross-platform transfer without retraining core visual backbones.
EmbodiedDiffusion: End-to-End Traversability-Guided Visual Diffusion for Heterogeneous Robot Navigation
arXiv:2512.02851v4 Announce Type: replace Abstract: Visual traversability estimation is central to autonomous navigation, yet most approaches either rely on prompt-driven Vision-Language Model (VLM) or decouple traversability from trajectory planning, requiring separate planners with heavy mapping, manual tuning, and extended deployment time. We propose EmbodiedDiffusion, a diffusion-based framework that simultaneously predicts traversability maps and generates feasible trajectories from RGB images using planner-free synthetic supervision and embodiment conditioning for cross-platform transfer. The framework distills category-level traversability semantics from a VLM teacher into a lightweight student model during training, enabling prompt-free, real-time inference at deployment. A modular FiLM-based conditioning mechanism isolates embodiment-specific reasoning into a compact trainable subset of the network, allowing rapid adaptation to new robot platforms without retraining the visual backbone or the trajectory diffusion model. Across indoor environments with quadruped and aerial robots, EmbodiedDiffusion achieves 80-100% navigation success in the full-data regime with real-time inference (90 ms) and adapts to new platforms using only 10 min of visual data collection, demonstrating scalable, unified traversability reasoning and trajectory generation for heterogeneous robots.
Abstract:Visual traversability estimation is central to autonomous navigation, yet most approaches either rely on prompt-driven Vision-Language Model (VLM) or decouple traversability from trajectory planning, requiring separate planners with heavy mapping, manual tuning, and extended deployment time. We propose EmbodiedDiffusion, a diffusion-based framework that simultaneously predicts traversability maps and generates feasible trajectories from RGB images using planner-free synthetic supervision and embodiment conditioning for cross-platform transfer. The framework distills category-level traversability semantics from a VLM teacher into a lightweight student model during training, enabling prompt-free, real-time inference at deployment. A modular FiLM-based conditioning mechanism isolates embodiment-specific reasoning into a compact trainable subset of the network, allowing rapid adaptation to new robot platforms without retraining the visual backbone or the trajectory diffusion model. Across indoor environments with quadruped and aerial robots, EmbodiedDiffusion achieves 80-100% navigation success in the full-data regime with real-time inference (90 ms) and adapts to new platforms using only 10 min of visual data collection, demonstrating scalable, unified traversability reasoning and trajectory generation for heterogeneous robots.
Robotics (cs.RO)
arXiv:2512.02851 [cs.RO]
(or arXiv:2512.02851v4 [cs.RO] for this version)
Text integral sub licența deschisă a publicației · 176 cuvinte
Subiecte conexe
EmbodiedDiffusion: End-to-End Traversability-Guided Visual Diffusion for Heterogeneous Robot Navigation
arXiv · 22 sept. 2026
Developing Healthcare Robotics with GPU-Native Medical Physics Simulation
NVIDIA · 28 iul. 2026
Learning Language-Conditioned Traversability Representations for Adaptive Visual Navigation
arXiv · 9 oct. 2026
AcousticDiffusion: Semantically Conditioned Audio-Guided Diffusion Policy for Search-and-Rescue Assistance
arXiv · 21 sept. 2026