Behavior Cloning for Active Perception with Low-Resolution Egocentric Vision
For roboticists, this shows that active perception can emerge from simple imitation learning without explicit reasoning, though the task is structured and the method is incremental.
Behavior cloning from low-resolution egocentric RGB images enables a robot arm to actively reposition its camera to center a partially visible plant before grasping, with relative joint delta prediction outperforming absolute position prediction.
We investigate whether behavior cloning is sufficient to produce active perception in a structured object-finding task. A low-cost robot arm equipped with a wrist-mounted egocentric RGB camera must reposition to center a partially visible plant before triggering a grasp signal, requiring actions that improve future observations. The model predicts joint commands directly from low-resolution RGB images under closed-loop control. We show that low-resolution egocentric vision is sufficient for reliable task completion and that predicting relative joint deltas substantially outperforms absolute joint position prediction in our setting. These results demonstrate that visually grounded active perception can emerge from behavior cloning in a reproducible setting.