Data‐Driven Modeling of Forces Exerted by Pneumatic Actuators for a Pediatric Exosuit

Accurate and interpretable models are essential for understanding actuator behavior and enabling effective design and control of soft wearable robotics. We present the experimental analysis and data‐driven modeling of soft pneumatic actuators designed to assist upper‐extremity motion in a pediatric exosuit. Experiments used an infant‐sized upper‐body test rig equipped with load cells and joint encoders to collect force and angle data. Four soft pneumatic actuators, two for the shoulder (1‐cell and 2‐cell rectangular actuators) and two for the elbow (bellow‐type actuators with square and circular cells), were selected based on workspace and motion profile smoothness. Each actuator was tested independently, with one joint actuated per trial across multiple anchoring configurations and passive joint angles to explore how these influence the actuated joint angles and generated forces. Three data‐driven approaches, segmented polynomial, segmented sigmoid, and autoregressive model with exogenous input (ARX), were used to model the relationship between pressure, joint angle, and force. The ARX model provided the most accurate and comprehensive representation by incorporating temporal information, whereas the sigmoid and polynomial models captured saturation and overall trends, respectively. These findings provide a basis for selecting modeling strategies suited to specific actuators and control needs, and support actuator design and model‐based control.