Engineered robotic fins have adapted principles of propulsion from bony-finned fish, using spatially-varying compliance and complex kinematics to produce and control the fin's propulsive force through time. While methods of force production are well understood, few models exist to predict the propulsive forces of a compliant, high degree of freedom, robotic fin as it moves through fluid. Inspired by evidence that the bluegill sunfish (Lepomis macrochirus) has bending sensation in its pectoral fins, the objective of this study is to understand how sensors distributed within a compliant robotic fin can be used to estimate and predict the fin's propulsive force. A biorobotic model of a bluegill sunfish pectoral fin was instrumented with pressure and bending sensors at multiple locations. Experiments with the robotic fin were executed that varied the swimming gait, flapping frequency, stroke phase, and fin stiffness to understand the forces and sensory measures that occur during swimming. A convolution-based, multi-input-single-output (MISO) model was selected to model and study the relationships between sensory data and propulsive force. Subsets of sensory data were studied to determine which sensor modalities and sensor placement locations resulted in the best force predictions. The propulsive forces of the fin were accurately predicted using the linear MISO model on intrinsic sensory data. Bending sensation was more effective than pressure sensation for predicting propulsive forces, and the importance of bending sensation was consistent with several results in biology and engineering studies. It was important to have a spatial distribution of sensors and multiple sensory modalities in order to predict forces across large changes to dynamics. The relationship between propulsive forces and intrinsic sensory measures is complex, and good models should allow for temporal lags between forces and sensory data, changes to the model within a fin stroke, and changes to the model through gait transitions.