Fish-Inspired Navigation via Flow Sensing in an Autonomous Robotic Swimmer

Autonomous ocean-exploring robots are challenged with navigating efficiently through complex, time-varying flow fields and seeking out areas of interest without prior knowledge of their surroundings. Aquatic animals, however, regularly accomplish this feat with a variety of flow sensing techniques. For example, fish are hypothesized to navigate by sensing velocity gradients with their lateral lines. Inspired by this navigation strategy, we placed distributed pressure sensors on a palm-sized robotic swimmer to mimic the function of canal neuromasts found in the lateral line of fishes, and tasked the robot with navigating efficiently through fluid flows in a 20-foot-tall water tank. Equipped with a high-speed microcontroller, the robot utilizes deep reinforcement learning and trains a neural network onboard in real time to control its actions. Additionally, we investigated thermal plume tracking with our robot platform, which has applications in tracking hydrothermal plumes and bio-signatures in the ocean.