A Hybrid Model and Learning-Based Force Estimation Framework for Surgical Robots

TL;DR

This work presents a hybrid model and learning-based framework for force estimation for the Patient Side Manipulators (PSM) of a da Vinci Research Kit (dVRK) and shows that by using a model-based method to perform dynamics identification, it reduces reliance on the training data covering the entire workspace.

Abstract

Haptic feedback to the surgeon during robotic surgery would enable safer and more immersive surgeries but estimating tissue interaction forces at the tips of robotically controlled surgical instruments has proven challenging. Few existing surgical robots can measure interaction forces directly and the additional sensor may limit the life of instruments. We present a hybrid model and learning-based framework for force estimation for the Patient Side Manipulators (PSM) of a da Vinci Research Kit (dVRK). The model-based component identifies the dynamic parameters of the robot and estimates free-space joint torque, while the learning-based component compensates for environmental factors, such as the additional torque caused by trocar interaction between the PSM instrument and the patient's body wall. We evaluate our method in an abdominal phantom and achieve an error in force estimation of under 10% normalized root-mean-squared error. We show that by using a model-based method to perform dynamics identification, we reduce reliance on the training data covering the entire workspace. Although originally developed for the dVRK, the proposed method is a generalizable framework for other compliant surgical robots. The code is available at https://github.com/vu-maple-lab/dvrk_force_estimation.

Figures (7)

• Figure 1: The Planar View of the Frame definition for the PSM. The unit of the link lengths is mm.
• Figure 2: The block diagram of Cartesian force estimation. We use the proposed model to estimate the torque required to achieve the measured joint positions and velocities with no contact. Then, we subtract it from the measured torque and use the Jacobian to convert the difference into force exerted on the environment.
• Figure 3: Experimental Setup. The instrument shaft is inserted through a trocar to the abdominal phantom cavity. The trocar is plugged into an incision port on the phantom, which is also the RCM of the robot in this configuration. The instrument tip manipulates the 3D-printed shaft and the force/torque sensor beneath it logs the force that the gripper exerts. Note that there is a 45-degree angle bias between the robot tip frame and the sensor frame, with respect to the sensor frame Z-axis.
• Figure 4: Joint position profile distribution of training and testing dataset collected in different workspace, which causes a mismatch.
• Figure 5: Learning-based vs. model-based free space torque estimation. Since there is a mismatch between training and testing datasets, the learning-based method has high error. Model-based method estimates well, as the dynamic parameters are identified following the proposed scheme.