A soft robotic approach for robot-assisted palpation training
File(s)
Author(s)
He, Liang
Type
Thesis
Abstract
This thesis investigates the coupled actuation and sensing in pneumatic-based soft robots, and further explores its applications in designing medical simulators for palpation training.
Palpation is a fundamental part of clinical training for physical examination. However, the learning approach remains a problem after centuries of modern medical development. A novel perspective to view this complex behaviour based on embodied physical intelligence pointed out that the learning of the technique ultimately relies on the training tasks and associated feedback. This opens up a new way to train medical practitioners with a quantitative archetype through soft robotic approaches.
Pneumatic-based soft robots with elastomeric material have the advantages of high-compliance, easy-for-control, and flexible-in-design. More importantly, the inherent soft sensing with the driving fluid shows promising sensitivity and reliability for human-robot interaction in palpation training.
Such an actuation/sensing approach was first employed in the design of a sensorized abdominal phantom with a controllable liver for realistic tumour rendering. Positive granular jamming nodules are used for tumour simulation instead of nodules made of pure rubber membrane. The method introduces extra constraints to the actuator where the volume/pressure curve can be effectively flattened. Thus, the valid control region is extended.
This thesis then explored the force and position sensing ability of the multi-nodule phantom with a machine-learning-based methodology. Reliable results are shown with a UR5 robot performing palpation with complex motion.
Finally, two applications of palpation training simulators based on the soft robotic approach were proposed: A full-sized abdominal phantom for physical environment training and a portable haptic interface for VR based training.
In summary, the findings from this thesis provide important notations for designing a pneumatic based soft robot with coupled actuation and sensing. In particular, such methodology shows great potential in developing medical simulators for high-fidelity haptic rendering that encapsulates tactile sensing.
Palpation is a fundamental part of clinical training for physical examination. However, the learning approach remains a problem after centuries of modern medical development. A novel perspective to view this complex behaviour based on embodied physical intelligence pointed out that the learning of the technique ultimately relies on the training tasks and associated feedback. This opens up a new way to train medical practitioners with a quantitative archetype through soft robotic approaches.
Pneumatic-based soft robots with elastomeric material have the advantages of high-compliance, easy-for-control, and flexible-in-design. More importantly, the inherent soft sensing with the driving fluid shows promising sensitivity and reliability for human-robot interaction in palpation training.
Such an actuation/sensing approach was first employed in the design of a sensorized abdominal phantom with a controllable liver for realistic tumour rendering. Positive granular jamming nodules are used for tumour simulation instead of nodules made of pure rubber membrane. The method introduces extra constraints to the actuator where the volume/pressure curve can be effectively flattened. Thus, the valid control region is extended.
This thesis then explored the force and position sensing ability of the multi-nodule phantom with a machine-learning-based methodology. Reliable results are shown with a UR5 robot performing palpation with complex motion.
Finally, two applications of palpation training simulators based on the soft robotic approach were proposed: A full-sized abdominal phantom for physical environment training and a portable haptic interface for VR based training.
In summary, the findings from this thesis provide important notations for designing a pneumatic based soft robot with coupled actuation and sensing. In particular, such methodology shows great potential in developing medical simulators for high-fidelity haptic rendering that encapsulates tactile sensing.
Version
Open Access
Date Issued
2021-01
Date Awarded
2021-05
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Nanayakkara, Thrishantha
Rojas Libreros, Nicolas
Publisher Department
Dyson School of Design Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)