Science1 publisher2 min readPublished
Bumblebees follow and face a swaying flower using the same motion cues in different ways
Wageningen and Delft researchers reproduced bumblebee landings on a flower model swaying twice a second with two control rules fed by the same cues. The pair gives engineers a control scheme to try on drones that must land on moving pads.
The Scientist · Science desk
What happened
- Three high-speed cameras filmed 48 landings, and the team rebuilt each flight in 3D to separate following the platform sideways from turning to face it.
- The bees matched the platform's sideways motion closely but trailed it in time, reacting to what they saw without having to predict the flower's path.
- To move sideways the bees banked, tilting to redirect their aerodynamic force much as a helicopter does.
- Taking the flower's offset from the center of view out of the turning model barely changed the modelled bees' performance, and motion cues mattered more.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- decision Anyone copying the scheme still has to choose whether to measure turn rate mechanically or visually, because the bee trials could not show which one the insects rely on.
- constraint The models describe a possible strategy without identifying the bees' sensors, so a drone that lands well with them would show the rules are sufficient, not that bees actually use them.
- contradiction A 2025 study from some of the same group linked the approach to keeping slower-swaying flowers centered in view. Set beside the new fit, how much the position cue counts looks tied to how the flower moves.
A flower's image can drift across a bee's view for three reasons: the flower moved, the bee moved, or the bee turned. The bee's reaction to the image does not show which of these happened [7]. Chenyao Wang's team at Wageningen University and Delft University of Technology put the problem where the bees had to solve it anyway. The swaying platform sat at the hive entrance, and landing on it was how the bees got home [1].
The platform travelled 2.5 cm out and back on each side of center, 10 cm per cycle. At two cycles a second, that is an average sideways speed of about 20 cm per second [2][1]. The banking result, in which tilt closely matched the acceleration the bees needed, comes from the 10 longest recordings [6]. That is about a fifth of the 48 landings filmed [2].
The models, published in the Journal of the Royal Society Interface [4], take two inputs. One is optic flow, the flower's apparent motion across the bee's view. The other is the bee's own turning rate [8]. Accounting for the turn removes the image change caused by rotation and leaves the motion caused by the bee sliding sideways relative to the flower [8]. Sideways tracking needed only a proportional rule: a larger combined signal produced a proportionally stronger acceleration [9]. Turning needed one more piece, a response to how quickly the optic flow was changing [10]. A control engineer would call that a derivative term added to a proportional one [10].
Simulations help explain why the flower's position in view added so little. A simulated bee that began its approach without rotating still turned toward the target, driven only by its responses to the flower's visual motion [15]. Under the tested conditions, those responses kept the flower centered without any separate correction for how far off-center it was [15]. So seeing a bee face a flower does not, by itself, tell you what information steered the turn [16].
The report does not give the size of the bees' lag behind the platform. For a drone running a purely reactive rule, that delay is the first figure I would want.
What to watch
- A flying-robot test of the two rules on a moving landing pad, showing whether a fit to 48 bee landings works as a controller.
- Bee trials with gust-like, unpredictable sway, where the researchers suggest the flower's position in view could become more useful.