Control of an Active Sensor Skin for Extreme Terrain Mobility

Robots made from soft materials can be used to assist in the exploration of planets, asteroids, or other bodies in space. The lightweight materials these robots are manufactured from could reduce the cost of transport, as well as increase the robustness of the robots to impacts, such as falling down hills or into craters. As the field of soft robotics is relatively new, much focus has been placed on developing materials, designs, and controls. While rigid robots have many existing sensors, such as encoders, IMUs, etc., that can be used to track joint positions, soft robots do not have equivalent sensors. Tracking the position of soft robots is challenging due to the lack of clearly defined joints and limited soft sensor choices. The goal of this research was to advance the field of soft robotics in terms of sensor-enabled proprioception and control through the use of active sensor skins. Active sensor skins are planar substrates with integrated sensors and actuators that can be wrapped around soft, passive structures to roboticize those structures. To achieve this goal, it was necessary to improve the robustness of current sensors to facilitate integration into a robotic system, as well as develop analytical models of systems that could assist in mapping between sensor outputs and the system state. Due to the complexity of this problem, focus was given to continuum arm-type robots, which are long, elephant trunk-like robots that have a strong foundation in modeling but had little work on inclusion of sensors for proprioceptive feedback. Most control work on continuum arms used either off-board sensing, such as cameras, or electromagnetic sensing, which can be problematic in certain environments. At the end of this research, we were able to improve the robustness of soft material strain sensors, which could be integrated into the active sensor skins, and develop analytical models that enabled the mapping of sensor responses to system state. The improved sensors used exfoliated graphite composite in a capacitive sensor. Exfoliated graphite was chosen since it does not significantly impact the material properties of the base material like other conductive particles, such as carbon black, would. The sensor was made to be capacitive since the sensor signal does not degrade with repeated use like it would have if the sensor been purely resistive. In order to enable proprioception in soft continuum robots, existing continuum arm models in literature were expanded to take input from strain sensors along the outside of the body to determine the state of the underlying structure. Further adaption of these models enabled the active sensory skins to determine the bending stiffness of the underlying structure, which can be valuable for ad-hoc development of feedforward control models. The ability to measure and understand the underlying structure can help improve controller performance if the active sensor skins are being used in remote locations, such as space. Additionally, these models were used to create a simulation environment which would help in the further development and design of future active sensor skins. The simulation environment also provides a testbed for new controller designs before testing on a physical system.