AxisGuide: Grounding Robot Action Coordinate System in RGB Observations for Robust Visuomotor Manipulation
Researchers have developed AxisGuide, a new method to improve robot visuomotor manipulation by grounding action coordinate systems in visual observations. This technique renders robot base-frame axes in camera views, providing explicit cues for motion in image space. Experiments show AxisGuide enhances performance and generalization for robots in both simulated and real-world tasks, particularly under distribution shifts. AI
IMPACT Enhances robot generalization and robustness in manipulation tasks by improving action understanding from visual input.