Three attempts to reduce natural phenomena to equations
- IRPHE, Marseille, invited by Rosa COSSART
During this talk, I will discuss three unrelated projects that share the curiosity-driven objective: can we better understand complex natural phenomena through physical and mathematical modeling?
- Trees shaped by the wind. Trees are self-similar branching structures, hierarchically organized with longer and thicker branches near the roots. With a mechanically-based numerical model, I will show how this self-similarity emerge through natural selection. In this model, trees grow into fractal structures to promote efficient photosynthesis and branch diameters increase in response to wind-induced loads. Remarkably, the virtual trees emerging from this model have self-similar properties similar to real trees.
- Form and function in fish swimming. Undulatory swimming is used by most aquatic animals. To achieve propulsion, these animals propagate a bending wave along their backbone down to their caudal fins. What is the optimal design for such an undulatory swimmer? To address this question, I will use a mechanical model and a genetic algorithm to calculate the minimal energetic costs, the maximal swimming velocity or any trade-off between the two.
- Tracking odor in turbulence. Life is tough for copepods! These millimetric crustaceans have about 1000 neurons at their disposal. Yet they are able to locate their mate by tracking the pheromone trail that they leave. That is no small feat for animals usually living in a turbulent environment and constantly washed by the flow. How are they able to do that? I will propose a way to reverse-engineer this problem with reinforcement learning tools.