Optimizing Base Placement of Surgical Robot: Kinematics Data-Driven Approach by Analyzing Working Pattern

TL;DR

This work addresses RAMIS base pose optimization by leveraging operator-specific working patterns inferred from end-effector kinematics. It introduces two kinematic scores, $score_{JM}$ and $score_M$, and trains an MLP regressor to map pattern-derived scores to the optimal continuous base pose. Validation in a dVRK-based simulation reveals operator-specific base-pose score maps and a substantial improvement over random base placements across unseen tasks, highlighting the value of tailoring base pose to individual operators. The framework is extensible to additional scores and can enhance RAMIS efficiency by embedding operator-specific preferences into preoperative planning.

Abstract

In robot-assisted minimally invasive surgery (RAMIS), optimal placement of the surgical robot base is crucial for successful surgery. Improper placement can hinder performance because of manipulator limitations and inaccessible workspaces. Conventional base placement relies on the experience of trained medical staff. This study proposes a novel method for determining the optimal base pose based on the surgeon's working pattern. The proposed method analyzes recorded end-effector poses using a machine learning-based clustering technique to identify key positions and orientations preferred by the surgeon. We introduce two scoring metrics to address the joint limit and singularity issues: joint margin and manipulability scores. We then train a multi-layer perceptron regressor to predict the optimal base pose based on these scores. Evaluation in a simulated environment using the da Vinci Research Kit shows unique base pose score maps for four volunteers, highlighting the individuality of the working patterns. Results comparing with 20,000 randomly selected base poses suggest that the score obtained using the proposed method is 28.2% higher than that obtained by random base placement. These results emphasize the need for operator-specific optimization during base placement in RAMIS.

Figures (7)

• Figure 1: Optimizing the robot base pose using the operator’s working pattern analysis. Kinematics Data: To analyze the operator’s working pattern (e.g., needle handling and grasping strategies), we use the recorded end-effector pose data from the previous operation. Working Pattern Analysis: We can identify representative end-effector poses by analyzing the poses frequently adopted by the end-effector. Score and Regression: We define two kinematics metrics to calculate the scores for the base pose. We perform multilayer perceptron (MLP)-based regression to determine the optimal base pose.
• Figure 2: Working Pattern Observation using the JIGSAWS dataset: Among the JIGSAWS dataset jigsaw, we selected two operators with expertise ($>$ 100 hrs) in operating surgical robots. Each operator exhibited different manipulation techniques when performing the suturing task (needle insertion, handover, needle extraction). These figures indicate that each operator has a unique working pattern.
• Figure 3: Simulation setup. (a) dVRK laparoscopic surgery training simulation setup. (b) Pick and place. (c) Peg on board. (d) Needle threading.
• Figure 4: Visited voxel and orientation clustering result. The first column shows the scattered positions of the voxels visited by the end-effector during surgery. Sequential IDs are assigned to each voxel to analyze the working pattern. The remaining columns display the clustering orientation results for the commonly visited voxels. The black arrow indicates the representative orientations, which are the centroids of each orientation cluster, expressed as rotation vectors. Interestingly, each volunteer explored different sets of voxels and adopted his or her preferred orientations even within commonly visited areas.
• Figure 5: Working pattern analysis : position analysis result. This is the analysis result of the working pattern using the end-effector’s position for the two volunteers mentioned in Fig. 4. The graph in the first row compares the visit counts for the five voxels most frequently visited by volunteer 1, whereas the graph in the second row compares the visit counts for the five voxels most frequently visited by volunteer 2.