Motion Planning and Constraint Exploration for Robotic Surgery

Abstract

Robot-assisted Minimally Invasive Surgery (MIS) presents patients with the benefits of reduced pain and shortened recovery time at a cost of perception deficiency to surgeons. In contrast to MIS, surgeons manually manipulate organs during open surgery and intuitively discern safe movements for flexibly constrained organs. This constraint perception ability is lost during robotic-assisted surgery and surgeons are expected to guard organs against trauma using other sensory cues such as vision feedback about tissue stretch. The aim of this research is to investigate algorithms that, when combined with future telemanipulation schemes would allow surgical robots to help surgeons with safe manipulation of organs. To achieve this goal we break new ground in the area of robotic manipulation and constraint exploration in flexible environments. The thesis focuses on three aspects of manipulation in flexible environments: 1) online path planning for safe manipulation of a flexibly constrained organ to a target pose, 2) global motion constraint estimation and characterization, 3) constraint identification and classification. Algorithms to estimate perceived flexible suspension constraint are developed based on estimation of the suspension stiffness tensor. Properties of the stiffness tensor are used to obtain frame-invariant principal constraint axes. The principal stiffness in this frame invariant representation is then used to form a 6-dimensional constraint vector that presents a compact characterization of perceived translational and rotational stiffness magnitudes in the principal axes directions. A method for compact characterization of global stiffness, as well as real-time mapping, classification and identification of constraints are then presented. Finally, Constraints are automatically defined and classified from local stiffness measurements, using a clustering algorithm.