Robotic pollination

Pollination is a rare case in field robotics where the biological incumbent is not merely cheap but cheaper by orders of magnitude. Both the problem statement and the real market follow from that.

What the insect is actually doing

For many crops, moving pollen is not the hard part. Tomato, aubergine, blueberry, cranberry and their relatives have poricidal anthers: the pollen is sealed inside a tube and leaves only through an apical pore. There is one way to release it, and that is vibration. A bumblebee grips the anther and contracts its thoracic flight muscles without moving its wings, producing an audible burst with a fundamental frequency of several hundred hertz; the pollen is ejected through the pore like powder from a shaken salt cellar. A honeybee cannot do this — hence the complete dependence of greenhouse tomato on bumblebee colonies, and the hand-held vibrating wands that preceded them.

The first engineering consequence is that a mechanical substitute must reproduce a frequency-and-amplitude regime, not a touch. Too little and the anther does not open; too much and the flower is damaged or the truss is knocked. Existing approaches divide into direct vibration (contact with the truss or the trellis wire), a directed air pulse, and precision deposition of pollen collected beforehand — the last used in orchards, where cross-pollination between cultivars is obligatory and pollen harvesting can be a separate operation.

Detection and force

The second difficulty is vision. A flower is small, often occluded by leaves, moving under its own weight and the ventilation, and it must not merely be found but staged: acting on an unopened bud is useless and on a spent flower harmful. The stigma’s receptive window is measured in days, so the machine has to revisit the same row repeatedly and remember what it has already treated. Positioning tolerance is centimetres while permissible force is hundredths of a newton — a manipulator working in the regime where ordinary stiff industrial kinematics does not apply.

The comparison is per flower

The honest metric is the cost of one successfully pollinated flower. A greenhouse bumblebee colony costs tens of euros, lives for some weeks and services thousands of square metres; each forager visits hundreds of flowers per trip, runs on no electricity, finds the flowers itself and distributes itself over the area. A robot must approach every flower individually. So these machines do not beat the bumblebee in open comparison, and do not try: they win where the insect is unavailable or not permitted.

There are three such niches, and they are real. Sealed greenhouses in countries that restrict bumblebee imports on biosecurity grounds. Breeding and seed production, where a specified parental combination is required and random transfer is unacceptable. And orchard crops with a short, weather-dependent bloom, where a cold or wet spell suppresses insect flight during precisely the days the stigma is receptive — there the machine competes not with a bee but with the absence of a crop.