Estimation in Minimally Invasive Surgical Robotics: Enabling Novel Technologies in Image Guidance and Sensing
AUTHORS
ABSTRACT
The overarching goal of this dissertation is to improve robotic, minimally invasive surgery with image guidance and advanced sensing techniques. Both of these approaches can provide valuable information to a surgeon to ultimately decrease patient trauma. In this dissertation, we apply tools from estimation theory to enable these novel technologies. Robotic surgery increases surgeon dexterity on its own; however, surgeon awareness is currently based solely on endoscope video. Surgical image guidance can improve surgeon awareness: the robotic instruments are displayed alongside 3D patient models within the surgeon console. Such instrument display involves computing the locations of the tools in real time–image guidance is only as accurate as the robot itself. Therefore, before use in image guidance, the accuracy of robot should first be evaluated. Chapter 2 focuses on this accuracy estimation (and improvement) for the two most common platforms. Chapter 3 presents an example image guidance system enabled by this work. Models can provide accurate tool positions in free space; however, once the surgeon uses the tools, applied forces can degrade model accuracy (e.g. shaft deflection, cable backlash). To address this, we propose to attach a tiny Inertial Measurement Unit (IMU) to the robotic instrument. By fusing kinematics data with IMU data, loading effects could be estimated online. This approach naturally presents a calibration problem: the alignment of the IMU chip on the instrument is initially unknown. Chapter 4 addresses and extends this problem. When access to surgical targets is nontrivial, continuum robots can produce smoothly curving shapes which closely match patient anatomy. Models can predict the shape of continuum robots given actuator values; however, shape is fundamentally coupled to interaction forces applied continuously along the robot’s backbone. Furthermore, it would be valuable to know both the location and magnitude of such contact with patient anatomy. In Chapter 5, we show how to estimate robot shape and distributed contact forces simultaneously by applying a powerful state estimation paradigm called continuous-time batch estimation to a continuum robot’s continuous-arclength domain.
Tags: Dissertations 2022