Development of a proprioceptive sensing for soft continuum robots
File(s)
Author(s)
Alian, Amirhosein
Type
Thesis
Abstract
Over recent decades, the incidence of early-onset Gastrointestinal (GI) cancer has risen considerably, highlighting the importance of regular screening. Conventional endoscopy is inherently invasive, requiring both patient sedation and highly skilled operators. Robotic flexible endoscopy offers enhanced precision, manoeuvrability, and procedural efficiency. Soft endoscopes are promising due to their dexterity and ability to adapt to the complex curvilinear geometry of the GI tract.
The intrinsic nonlinearity of soft robots poses significant challenges for control, primarily due to the lack of compatible state estimation mechanisms. Existing sensing technologies suffer from limitations in scalability, integrability, cost, and biocompatibility, particularly for endoscopic applications. This thesis explored the development and validation of an Electrical Impedance Tomography (EIT)-based sensing methodology for Soft Fluidic Actuators (SFAs), aimed at advancing proprioception and exteroception for lower GI soft continuum endoscopes.
Initially, the concept was validated through simulations and planar experiments. A Flexible Printed Circuit (FPC) array was embedded into a Soft Continuum Manipulator (SCM), allowing distributed impedance measurements. A Long Short-Term Memory (LSTM) model reconstructed the manipulator’s planar tip displacement, confirming the feasibility of data-driven shape estimation.
The EIT framework was extended to exteroception through tactile interaction studies. A dual-mode sensing approach enabled the estimation of contact magnitude, with further translation into soft tissue phantom palpation.
In 3D, EIT, through a stretchable FPC, and a Multi-Layer Perceptron (MLP) network achieved an average Root Mean Square Error (RMSE) of 0.37 mm. Further validation in clinical scenarios achieved an RMSE of 1.34 mm under ex vivo conditions, followed by an in vivo trial in a 35 kg porcine model, where usability, robustness, and safety were
evaluated.
Overall, this work established EIT as a compact, scalable, and biocompatible perception for soft continuum endoscopes. The findings provided a foundation for future research in transfer learning and EIT-based closed-loop control.
The intrinsic nonlinearity of soft robots poses significant challenges for control, primarily due to the lack of compatible state estimation mechanisms. Existing sensing technologies suffer from limitations in scalability, integrability, cost, and biocompatibility, particularly for endoscopic applications. This thesis explored the development and validation of an Electrical Impedance Tomography (EIT)-based sensing methodology for Soft Fluidic Actuators (SFAs), aimed at advancing proprioception and exteroception for lower GI soft continuum endoscopes.
Initially, the concept was validated through simulations and planar experiments. A Flexible Printed Circuit (FPC) array was embedded into a Soft Continuum Manipulator (SCM), allowing distributed impedance measurements. A Long Short-Term Memory (LSTM) model reconstructed the manipulator’s planar tip displacement, confirming the feasibility of data-driven shape estimation.
The EIT framework was extended to exteroception through tactile interaction studies. A dual-mode sensing approach enabled the estimation of contact magnitude, with further translation into soft tissue phantom palpation.
In 3D, EIT, through a stretchable FPC, and a Multi-Layer Perceptron (MLP) network achieved an average Root Mean Square Error (RMSE) of 0.37 mm. Further validation in clinical scenarios achieved an RMSE of 1.34 mm under ex vivo conditions, followed by an in vivo trial in a 35 kg porcine model, where usability, robustness, and safety were
evaluated.
Overall, this work established EIT as a compact, scalable, and biocompatible perception for soft continuum endoscopes. The findings provided a foundation for future research in transfer learning and EIT-based closed-loop control.
Version
Open Access
Date Issued
2025-11-17
Date Awarded
2026-03-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Mylonas, George
Avery, James
Rodriguez Y Baena, Ferdinando
Sponsor
Imperial College London
Publisher Department
Department of Surgery & Cancer
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)