Efficient point-to-point navigation in the presence of a background flow field is important for robotic applications such as ocean surveying. In such applications, robots may only have knowledge of their immediate surroundings or be faced with time-varying currents, which limits the use of optimal control techniques. Here, we apply a recently introduced Reinforcement Learning algorithm to discover time-efficient navigation policies to steer a fixed-speed swimmer through unsteady two-dimensional flow fields. The algorithm entails inputting environmental cues into a deep neural network that determines the swimmer’s actions, and deploying Remember and Forget Experience Replay. We find that the resulting swimmers successfully exploit the background flow to reach the target, but that this success depends on the sensed environmental cue. Surprisingly, a velocity sensing approach significantly outperformed a bio-mimetic vorticity sensing approach, and achieved a near 100% success rate in reaching the target locations while approaching the time-efficiency of optimal navigation trajectories.

Fish must maneuver laterally to maintain their position in schools or near solid boundaries. Unsteady hydrodynamic models, such as the Theodorsen and Garrick models, predict forces on tethered oscillating hydrofoils aligned with the incoming flow. How well these models predict forces when bio-inspired hydrofoils are free to move laterally or when angled relative to the incoming flow is unclear. We tested the ability of five linear models to predict a small lateral adjustment made by a hydrofoil undergoing biased pitch oscillations. We compared the models to water channel tests in which air bushings gave a rigid pitching hydrofoil lateral freedom. What we found is that even with no fitted coefficients, linear models predict some features of the lateral response, particularly high frequency features like the amplitude and phase of passive heave oscillations. To predict low frequency features of the response, such as overshoot and settling time, we needed a semiempirical model based on tethered force measurements. Our results suggest that fish and fish-inspired vehicles could use linear models for some aspects of lateral station-keeping, but would need nonlinear or semiempirical wake models for more advanced maneuvers.