Estimation of the existence of flight policies in echolocating bats

Echolocating bats routinely navigate complex natural environments with remarkable agility, yet it remains unclear whether their flight trajectories reflect reproducible internal control strategies or are simply improvised reactive responses. To resolve this question, we recorded the flight paths and ultrasonic pulse emissions of Rhinolophus nippon and Miniopterus fuliginosus as they negotiated seven obstacle-rich arenas in complete darkness. We then trained a machine-learning model that, after observing only the initial segment of each trajectory, accurately predicted the animals’ subsequent paths across arenas and individuals, faithfully preserving features such as turning direction, obstacle-avoidance arcs, and velocity profiles. The model’s ability to replicate species-specific strategies despite pronounced differences in sonar systems and flight morphology indicates that bat navigation is governed by structured internal control policies. Our findings demonstrate that bat flight, while seemingly improvisational, is in fact steered by consistent internal rules that can be inferred directly from trajectory data without hand-crafted assumptions.