A Composite Soft Bending Actuation Module with Integrated Curvature Sensing Selim Ozel, Erik H. Skorina, Ming Luo, Weijia Tao, Fuchen Chen, Yixiao Pan, and Cagdas D. Onal Abstract— Soft robotics carries with the promise of making robots as capable and adaptable as biological creatures, but this will not be possible without the ability to perform self- sensing and control with precision and repeatability. In this paper, we seek to address this need with the development of a new pneumatically-actuated soft bending actuation module with integrated curvature sensing. We designed and fabricated two different versions of this module: one with a commercially available resistive flex sensor and the other with a magnetic curvature sensor of our own design, and used an external motion capture system to calibrate and verify the validity of these two modules. In addition, we used an iterative sliding mode controller to drive the modules through step curvature references to demonstrate the controllability of the modules as well as compare the usability of the two sensors. We found that the magnetic sensor returned noisy but accurate data, while the flex sensor was inaccurate and subject to drift but did not exhibit notable noise. Experimental results show that this phenomenon of drift from the flex sensor causes active feedback control of the bending actuator to exhibit significant positioning errors. This work demonstrates that our soft bending actuator can be controlled with repeatability and precision, and that our magnetic curvature sensor represents an improvement for use in closed-loop control of soft robotic devices. I. INTRODUCTION Pneumatically actuated soft robots have many exciting properties, but these properties are largely academic without the ability to perform self-sensing and control with precision and repeatability. The compliant nature of these robots, which is one of their strengths, stymies traditional efforts to sense their state due to the infinite passive degrees of freedom provided by flexible links. In addition, the dynamic behavior of soft actuators includes a nonlinear and non-trivial time delay as pressurized air is introduced through solenoid valve commands. Thus, new methods of on-board sensing and control need to be developed to allow soft robots to be used in real environments to solve practical problems. To address a lack of proprioceptive sensing in soft robotics research, this paper introduces a soft bending actuator mod- ule with embedded curvature measurements as a solution to sensing and control challenges related to soft robots. This is a step towards autonomous soft robots with self con- tained modules. Integrated sensing can be achieved through Hall effect elements or resistive flex sensors embedded in the constraint layer (neutral bending axis) for curvature measurement. Figure 1 displays the parts of the proposed segment. For segment control, we previously presented a The authors are with the Mechanical Engineering Department and Robotics Engineering Program, Worcester Polytechnic Institute, MA 01609, USA. All correspondence should be addressed to Cagdas D. Onal cdonal@wpi.edu Fig. 1. Isometric (a), right (b) and front (c) views of the sensor assembly are shown. It is composed of three distinct parts: a soft bending actuator, a constraint layer in the middle of (a) and a curvature sensor. Pictured here is our custom Hall Effect sensor, though the flex sensor is fit into the same layer. Circuit tracks for our sensor are etched on a flexible sheet. A hall effect and magnet pair is used for measurements and they can be seen in (c). hybrid-approach that used pulse width modulation (PWM) of valves to regulate pressure inside chambers [1]. Work in this paper utilizes this controller with embedded sensing to control curvature of a soft segment. Numerous novel motions that were previously not achiev- able by their rigid counterparts [2], [3], [4], [5], [6], [7] were recently demonstrated using soft-bodied robot designs. We believe the field is mature enough and requires a stronger understanding of controllers, sensors and actuators that would help soft robots to achieve an increased level of autonomy. Soft robots can have very different configurations depending on the design. Indeed one advantage of soft robots is the ability to achieve relatively complex motions such as quadruped locomotion through simple designs [8]. Neverthe- less, more complex tasks or locomotion in uncontrolled, open environments would eventually need feedback controllers. In this paper, we focus our attention on developing low- level curvature controllers and corresponding sensor systems appropriate for our recent soft robot designs in [9] and [10]. Accurate and embedded sensing along with a controller for regulating deformation is required for robots made out of soft bending actuators. Our snake robot in [9] is able to move without feedback control of its segments. Nevertheless, some tasks, such as finding its way through constrained environments and narrow spaces, would require some sort of feedback for motion control and planning. Visual tracking of robot configuration is a common practice in robotics. Specifically for soft-bodied robots, external motion capture may monitor the continuous kinematic configuration of the