Fig. 2: Training results for barotactic migration.
From: Leveraging agent-based models and deep reinforcement learning to predict taxis in cell migration
A Reward across episodes in the straight, top, and bottom geometries, with a zoomed-in detail of the first 500 episodes, reaching a mean value of 0.9999 after 7543 episodes. B Trajectories of the cell across the top, straight, and bottom geometries after the training process. C Mean normalized pressure around the cell in each geometry, with shaded regions indicating the minimum and maximum values.
