Real-time adaptive control of robots for microsurgery: towards submillimeter accuracy without added sensors
【研究キーワード】
手術ロボット / 適応制御 / 画像処理 / ロボティクス / センサ融合
【研究成果の概要】
The online calibration of surgical instruments with partial or complete task space measurements using external sensors was investigated.
(1) The proposed adaptive controller has been finalized. The adaptive controller can consider partial and complete task-space measurements and workspace constraints. Any type of sensor can be used.
(2) The proposed controller was validated in several experiments using the SmartArm system. First, using a high-cost high-accuracy image tracking sensor, online calibration using partial measurements was studied. It was shown that the adaptive controller can improve the accuracy of the model. In addition, in a novel, replicable experimental setup using a (comparatively) lower-cost camera system, the adaptive controller was evaluated in a scenario with nonlinear task-space constraints. It was shown that the model was able to adapt online for model inaccuracies and prevent task-space collisions.
(3) A method for pose tracking of rigid instruments has been developed using deep learning. Validation was done using real robotic rigid instrument images. An extension of this method was studied for instruments with multiple degrees of freedom.
(4) A framework for the generation of realistic CG images using VR simulation was developed. It has been of great use in the training of deep learning algorithms for instrument tracking.
The many contributions of this project have been submitted and/or published in top-level international conferences and peer-reviewed international journals.
【研究代表者】
【研究種目】若手研究
【研究期間】2019-04-01 - 2023-03-31
【配分額】4,290千円 (直接経費: 3,300千円、間接経費: 990千円)