Medical robots are increasingly adopting soft robotic technologies, to minimise the damage left behind by robotic interventions and explore new therapeutic treatments. For long-term implantable robots interacting with tissue, there are a plethora of challenges presented by the in vivo context, including biocompatibility, safety and compactness of design. Soft robots afford many advantages and answers to these contextual requirements, including their inherent compliance and hyperelasticity, providing the means for both increased safety, as well as generating large expansions from compact designs. However, the use of soft robots is limited by their infinite degrees of freedom, state-dependent and nonlinear behaviours, a strong dependence on the external environment, and difficulties in differentiating between stimuli. Soft sensors capable of giving information on the system and environment as a whole, as well as any abnormalities and faults, and, importantly, able to adapt to the changes in both the robot and the environment, will be key in driving, and securing, the future of soft technologies in medical robots.
In this thesis, a variety of sensing technologies are implemented and explored to advance the development of safer and more effective medical robots. Firstly, a soft sensor actuator, capable of large expansion and differentiating between height and pressure, is introduced as a building block for a variety of demonstrated probe and implant applications. Pressure and resistance sensing were shown to be interchangeable, although best used in combination, and a best estimated resolution of 4 g and 0.07 mm was achieved. The integration of these sensing technologies to distinguish between internal and external stimuli is developed further in the context of tissue lengthening soft robots. A three-columned sensor-actuator is shown to have good structural strength, shrinking by only 12 % under a 500 g load and growing more than 80 % of its unloaded height when preloaded with 250 g. The resistance sensing also proved useful for bending characterisation, but further work is still needed to improve repeatability. Finally, new methodologies, to maximise the information of localisation of stimuli and to tune the sensors to a robot and its context, are presented. The ability to tailor and tune sensing is preliminarily demonstrated as an opportunity to adjust sensing dynamically to the context. Additionally, exploiting the dynamics of fluid-based sensing is shown, allowing single and multiple press events and their locations to be determined.
