Estimating propulsive efficiency of bottlenose dolphins during steady-state swimming*

Cetaceans are phenomenal swimmers, but the marine environment makes it difficult to directly observe and quantify their dynamic swimming behavior. Biologging tags are often used to measure animal movement in the wild. But these embedded systems only measure movement kinematics where they are attached, and cannot measure the hydrodynamic forces the animals use to swim. Here, we present a framework that leverages a low-order model of dolphin swimming dynamics and kinematic data (orientation, depth, speed) collected from a biologging tag to: A) estimate the sagittal-plane body kinematics of swimming bottlenose dolphins (Tursiops truncatus); and B) estimate swimming kinetics and propulsive efficiency during steady-state swimming. Body kinematics for the segmented dolphin model were estimated from tag data using a temporal convolutional network that was trained using a synthetic data set. The estimated segment angles had errors of less than 2° from the true body joint angles. The measured and estimated kinematic data were used as inputs for the dolphin model to estimate the internal and external forces generated during swimming. The estimated kinematics and kinetics compare with published results, and the estimated propulsive efficiency were typically greater than 70% across the range of swimming speeds investigated. These results enable per stoke estimates of propulsive efficiency, and provide the foundation for an approach that can be used in the future to estimate the swimming biomechanics of dolphins in the wild.