Bee-Nav: Efficient robot navigation inspired by honeybee learning flights

Ajouté : 2026-05-20

[00:00:00]B-Nav We present B-Nav, [music] an extremely efficient robot navigation method inspired by honeybee learning flights. [music] When a robot first leaves home, it makes a short [music] learning flight nearby.

[00:00:18]>> [music] >> During this flight, a neural network learns to estimate the direction and distance home from images.

[00:00:29]After [music] learning, like a honeybee, the robot can immediately fly farther away.

[00:00:38]During this outward [music] flight, it tracks its position using odometry.

[00:00:45]After its task, the robot uses odometry [music] to head home, but drift means it may not return exactly to where it started.

[00:00:53]Once it reaches [music] the learned homing area, the neural network guides it the rest of the way home.

[00:01:00]>> [music] [music] [music] [music] >> B-Nav will enable [music] small robots to navigate over longer distances.

[00:01:57]This is useful for robots that travel out and then return [music] to a central home location, such as lightweight drones monitoring greenhouses.

[00:02:07]It also offers [music] a new perspective on how flying insects find their way home.

[00:02:13]>> [music] >> For example, our study suggests that learning with odometry produces winding paths home, which could be tested in biological [music] experiments.

[00:02:34]>> [music] [snorts] >> Oops.

[00:02:50]Help.

[00:02:52]>> [screaming]