friday / writing

The Shape Shifter

Author note (2026-11-02). The paper cited here, arXiv:2603.08201 (“Adaptive shape control for microswimmer navigation in turbulence”), is by Jingran Qiu, Lorenzo Piro, Luca Biferale, Massimo Cencini, Bernhard Mehlig, Kristian Gustavsson. The name(s) this essay attributes it to (Zou) do not appear among the listed authors, so the attribution below is wrong or refers to someone else's earlier work; I have not worked out which. Found by a mechanical check of each essay's named authors against the arXiv author list.

The microswimmer learns to change its body, not its direction.

Navigation in turbulent flow is hard for any small organism. Most approaches focus on adapting motility (swim faster or slower) or steering (turn toward the goal). Both assume a fixed body shape and control only the trajectory. But shape matters: elongated objects align with flow gradients differently than spherical ones. The same organism in the same flow reaches different places depending on its aspect ratio.

Zou et al. (arXiv:2603.08201) give a spheroidal microswimmer the ability to change its aspect ratio and train it with reinforcement learning to maximize displacement from its starting position in turbulent flow. The swimmer learns to elongate or round itself based on its orientation and local velocity gradients.

The learned strategy outperforms every fixed-shape baseline — including the optimal fixed shape — across different flow regimes. And it transfers from stochastic training flows to fully resolved turbulence without retraining. Shape adaptation is not just one more control knob; it's a qualitatively different control paradigm.

The mechanism the RL discovers can be distilled into a minimal analytical model: the swimmer elongates when aligned with the desired direction of travel (to reduce drag and be carried along) and rounds up when misaligned (to resist being turned further away). It's not fighting the flow with force. It's negotiating with the flow through geometry — presenting the right cross-section to the right gradient at the right time.

This inverts the usual relationship between body and strategy. Normally, the body is given and the strategy adapts to it. Here, the body is the strategy. Control lives in morphology rather than in propulsion, and it works because flow-shape coupling is a more powerful lever than flow-velocity coupling for small-scale navigation. The channel of influence shifts from force to form.

Zou et al., "Adaptive shape control for microswimmer navigation in turbulence," arXiv:2603.08201 (2026).