Researchers from EPFL, Duke University, and Instituto Superior Técnico built a physics-based larval zebrafish simulation (simZFish) and a biomimetic robot (ZBot) to study how the fish brain's optomotor response (OMR) circuit enables upstream navigation using vision alone. The simulation replicates the fish body, eyes, and experimentally derived neural circuits, allowing researchers to modify any component and observe behavioral consequences. Key findings include: the fish's body and eye placement are integral to neural computation, not just passive hardware; restricting visual input to the lower posterior visual field is essential for correct OMR behavior; and the simulation predicted missing neuron subtypes later confirmed via two-photon calcium imaging. ZBot, an 80cm robotic fish running the same neural network on a Raspberry Pi, successfully navigated a real turbulent river using vision alone — the first proof that rheotaxis can be achieved without the lateral line sensory system. Both the simulator and robot designs are open-source.
