Teleoperation of a surgical robot using force feedback
Authors
Krogh, Simon Bjerre ; Bolgár, Dániel ; Maric, Filip ; Silvani, Nicolas Pierre Michel
Term
1. term
Education
Publication year
2016
Submitted on
2016-12-20
Pages
6
Abstract
Denne afhandling undersøger, om kraftfeedback i teleoperation af kirurgiske robotter kan implementeres uden ekstra sensorer ved at udnytte eksisterende hardware på en da Vinci-platform. Formålet er at genskabe kirurgens følesans og øge kontroltransparensen ved at estimere de eksterne kræfter ved EndoWrist-værktøjet ud fra aktuatorernes målinger, frem for at måle kræfter direkte. Systemet bruger en Geomagic Touch som haptisk grænseflade til at styre EndoWristens fire frihedsgrader, mens en mellemliggende computer estimerer reaktionskræfter og håndterer kommunikationen med den indlejrede styreenhed. På grund af værktøjets ikke-lineære dynamik udvikles kraftestimatorer via systemidentifikation: rullemomentet modelleres særskilt, mens pitch- og yaw-kræfter identificeres som Hammerstein–Wiener-modeller med lineære tilstandsrums-kerner (6. orden) og dødzoner, baseret på aktuatorindsats og -hastighed. Tilbagemeldingen kortlægges til Geomagic Touch’ tre aktuatoriske akser (rul, klem-afledt yaw og klem-pitch). For at reducere latenstid erstattes en oprindelig TCP/JSON-strøm (~100 Hz) med UDP og kompakte pakker over direkte Ethernet. I den beskrevne opsætning leverer systemet en lukket feedback-løkke på cirka 500 Hz (mål: 550 Hz; arbejde mod op til 1 kHz), mens de identificerede modeller viser konsistent overensstemmelse med ubrugte målinger. Resultaterne indikerer, at en softwarebaseret, sensorløs kraftfeedback er praktisk mulig, men der kræves yderligere kvantitativ evaluering af nøjagtighed og stabilitet for fuld validering.
This thesis investigates whether force feedback for teleoperated surgical robots can be delivered without additional sensors by leveraging existing hardware on a da Vinci platform. The goal is to restore the surgeon’s sense of touch and maintain controller transparency by estimating external forces at the EndoWrist from actuator measurements rather than measuring forces directly. The system employs a Geomagic Touch haptic device to command the EndoWrist’s four degrees of freedom, with an intermediary computer estimating reaction forces and managing communication with the embedded controller. Because the tool exhibits strongly nonlinear dynamics, force estimators are built using system identification: roll torque is modeled separately, while pitch and yaw forces are identified as Hammerstein–Wiener models with linear state-space cores (6th order) and dead zones, based on actuator effort and velocity. Feedback is mapped to the three actuated axes of the Geomagic Touch (roll, clamp-induced yaw, and clamp pitch). To reduce latency, the original TCP/JSON stream (~100 Hz) is replaced with UDP and compact packets over direct Ethernet. In the presented setup, the closed feedback loop runs at approximately 500 Hz (target: 550 Hz; work toward up to 1 kHz), and the identified models qualitatively match previously unused measurements. These results suggest that a software-only, sensorless force-feedback approach is feasible, although further quantitative assessment of accuracy and stability is needed for full validation.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents
Other projects by the authors