UC Santa Cruz

Driven by the challenges in the adoption of state of the art autonomous capabilities in commercial surgical robot systems, our work aims to bridge the divide between prototypes and commercial adoption. In order to do this, there is a need to establish a systematic way of evaluating safety and effectiveness of autonomous capabilities. One reflection of safety and effectiveness is the interplay between robot capability and task complexity. The complexity-capability relation is a crucial part of robot performance and has the potential to dictate surgical outcomes. In this dissertation, we explored and presented metrics that surfaced these relations. Two problems influenced by the complexity-capability relations were studied: first, we look at task pose planning problem in order to enable safe and reliable execution of automated sub-tasks; second, we explored the problem of quantifying robot-task compatibility to facilitate comparison of automated task executions.

For the task pose planning problem, we considered suture looping task pose planning in robot-assisted minimally invasive surgery (RAMIS) and bone pose identification in robot-assisted orthopedic surgery (RAOS). For the suture looping task pose planning, we proposed a combined linear programming for task volume position optimization and a brute force orientation search method to locate a robustly safe task pose. The planner was extensively verified in simulation to effectively avoid instrument-tissue collisions, suture entanglements and effective length constraints, and gripper joint limits. For bone pose identification in RAOS, we proposed a task-specific capability map that captures the capability of the robot to execute a bone cutting task at various task poses. A feasibility measure that accounts for patient anatomy and user bone placement error was developed for searching suitable bone poses. The method was validated using TCAT, a 5 degree-of-freedom (DOF) surgical robot, to be effective in providing anatomically feasible bone poses.

In quantifying robot and task compatibility, we proposed the error rate compatibility measure. Unlike existing Jacobian based measures where compatibility evaluation is connected to robot performance, the proposed measure is based on the number of simple motion approximation when translating task space path to joint space. This approach makes the measure scale and unit independent and overcomes some of the inherent limitations of Jacobian based measures. Numerical simulation results for 3- and 6-DOF manipulators showed that it is effective in differentiating compatibility when comparing robot to robot and task to task designs.