Design and Validation of a Multimodal Dataset of Robot-Assisted Suturing Gestures based on Kinematic and Force Information

Surgical activity knowledge is a fundamental skill for the automation of robot-assisted surgical interventions. Surgical robots need not only to recognize kinematic patterns that constitute the surgical tasks but also to provide physical interpretation to understand the surgical workflow. Data are essential for training robots to recognize the procedure being performed and eventually automatize it with a customized control strategy based on the specific sub-task. Moreover, the interaction of the robot with the environment plays a crucial role in safely performing complex tasks and operating with greater autonomy and precision. The open access design of the da Vinci Research Kit (dVRK) control architecture incentivizes and enables researchers to easily access and collect data. For these reasons, this work proposes a novel dataset based on the suturing task on dVRK from several surgeons with different skill levels. The dataset includes kinematics and interaction force data labeled with the surgical gestures part of the suture procedure.