The proposed framework addresses the challenge of providing situational awareness in robotic systems for manipulation tasks under microscopic view. It mounts a camera on a robotic arm to enable its controllable motion, ensuring the field-of-view (FoV) remains focused on a tool being manipulated.
The key aspects of the framework are:
Modeling the camera extrinsics as part of the overall kinematic model of the robotic system. This allows the framework to adaptively update the estimated model parameters, including the camera pose, using real-time feedback from a marker-less tool tracking algorithm.
Formulating an optimization-based control strategy that simultaneously controls the robot holding the tool to follow a desired trajectory, while also autonomously moving the camera-holding robot to maintain the tool within the camera's FoV. Workspace constraints, including collision avoidance and FoV limits, are incorporated into the control problem.
Evaluating the proposed framework in a proof-of-concept bi-manual robotic setup, where a microscopic camera is controlled to view a tool moving along a pre-defined trajectory. The adaptive control strategy allowed the camera to stay within the real FoV 94.1% of the time, compared to only 54.4% without adaptation.
The framework enables robust autonomous control of a robot-mounted camera to provide consistent visual feedback during microscale manipulation tasks, overcoming the challenges posed by the limited FoV of high-magnification cameras.
לשפה אחרת
מתוכן המקור
arxiv.org
תובנות מפתח מזוקקות מ:
by Hung-Ching L... ב- arxiv.org 04-01-2024
https://arxiv.org/pdf/2309.10287.pdfשאלות מעמיקות