The Windhover

A kestrel hunts by hanging motionless in a gusty wind — body thrashing, wings sawing, but its head locked in space to within millimetres. It's a two-loop biological gimbal, and spacecraft engineers are copying it.
after Necker 2006 (birds have two vestibular organs) · ESA ACT bird-inspired stabilization · Hopkins, 1877
side view — body buffeted by wind, head held on target
head RMS error: °  ·  stabilization gain: ×
what the head sees (gaze on prey)
pitch traces — body vs head

How it works. Wind gusts hit the body as random torque — it pitches and heaves. A cascaded controller drives the neck: an outer loop (the vestibulocollic reflex, sensing head-in-space) commands the neck angle to cancel body motion, and a fast inner loop (the vestibulo-ocular reflex, ~10 ms) trims the eye on top. The head becomes a stabilized sensor platform floating above a chaotic base — exactly the problem a camera gimbal solves, except the kestrel does it with two vestibular organs and a very flexible neck. Toggle the reflex off and watch the gaze smear.

The demo is a real controller: body dynamics are a damped rotational mass driven by band-limited gust noise; the neck is a PD loop with sensor latency and a rate limit (necks aren't infinitely fast). The × number is the ratio of body-pitch RMS to head-pitch RMS — how many times more steady the gaze is than the body carrying it.

a break, well spent · disturbance rejection is disturbance rejection, whether it's a neck or an α-vector