LMP Seminar: Performance of smart microswimmers in a turbulent flow in the presence of obstacle using Q-learning strategies

Motility has played a crucial role in the evolutionary success of microorganisms, enabling them to adapt to their environment more effectively. This adaptive strategy is not limited to navigating complex environments to reach nutrition but also in avoiding obstacles; like sperm transport, where millions of spermatozoa face a series of obstacles within varying fluid environments. Using head-based chemotaxis, sperm cells adjust their propulsion in response to environmental chemical gradients to enhance their chances of successful fertilization. Similarly, bacteria like E. coli employ a "run and tumble" strategy to explore their surroundings, such as the digestive tract, where they assess their proximity to attractants like sugars and amino acids. When a repellent is detected, E. coli increases the frequency of tumbles, changing direction to move away from harmful stimuli. Although swimmer navigation in steady flows has been well-documented and studied, integrating these principles into models for obstacle-laden environments remains an open challenge. Reinforcement Learning (RL) in multiagent systems involves multiple agents learning to interact with an environment to achieve individual or collective goals. Using RL, I was able to show that smart microswimmers in turbulent flows in the presence of an obstacle can surpass in number as compared to naive swimmers. I also performed transfer learning with trained microswimmer Q matrix and compared them to surfers.