@@ -356,13 +356,20 @@ \subsubsection*{Bayesian Optimization}
356356
357357
358358% ###################################################################################################
359- \section {Target System }
359+ \section {Target Systems }
360360\label {section:target-system }
361361\begin {itemize }
362362 \item Make clear that black-box optimization is required
363363 \item Should have some application for real-world scenarios
364364\end {itemize }
365365
366+ In order to compare various optimization methods, we study two distinct target systems.
367+ The first one aims to estimate parameters of the \ac {abm} by comparing individual agents with
368+ experimental data.
369+ Meanwhile, the second system represents the use-case of an objective function which is applied to a
370+ large collection of agents.
371+ We use cellular\_ raza~\cite {Pleyer2025 } to implement the numerical realization of these systems.
372+
366373% ---------------------------------------------------------------------------------------------------
367374\subsection {Bacterial Rods }
368375\label {subsection:bacterial-rods }
@@ -373,6 +380,41 @@ \subsection{Bacterial Rods}
373380 \item neighbor interaction $ \rightarrow $ not diff.-able
374381\end {itemize }
375382
383+ This model describes the growth and interactions of rod-shaped bacteria such as \textit {E.Coli } or
384+ \textit {B.Subtilis }.
385+ It has been published with cellular\_ raza~\cite {Pleyer2025 } and consists of 4 simulation aspects.
386+ \paragraph {Mechanics }
387+ individual bacteria are represented by a collection of vertices $ \vec {v}_i$ which are coupled to
388+ each other via springs.
389+ In addition, a bending force acts between the angles of connecting edges, thus promoting a straight
390+ shape without curvature.
391+ \paragraph {Interaction }
392+ Bacteria exert forces which repell or attract each other.
393+ These forces act between a single vertex $ \vec {v}_i$ on one agent and the closest point on each
394+ connection between two vertices $ \vec {w}_j$ and $ \vec {w}_{j+1}$ of the interacting agent.
395+ \paragraph {Cell Cycle }
396+ Each cell-agent is continuously growing, thus extending the length of the rod via the insertion of
397+ new material at the cylindtrical part.
398+ Once they reach a certain length threshold, the bacteria divide in the middle.
399+ During this process, new vertices are determined such that the resulting two new bacteria align
400+ identically within the path of the vertices of the mother-cell.
401+ \paragraph {Domain }
402+ We can choose the simulation dimension (2D,3D).
403+ For the estimation of individual parameters, we restrict ourselves to 2D to realistically depict the
404+ bacterial growth on a plate.
405+ The domain has reflective boundary conditions which are however not applied since our bacteria are
406+ located far enough in the center of the simulation domain.
407+
408+ Figure~\ref {fig:bacterial-rods-sim } shows various results of numerical simulations of this model.
409+ Subfigure (A-B) show the growth of these bacteria inside a cuboid, thus slowly filling up the space
410+ from the left to right and favoring an alignment of the rods along this dimension of growth.
411+ Meanwhile, subfigures (C-E) show snapshots of a shallow 3D simulation where the postiion of agents
412+ has been converted to individual cell-masks such as produced by cell-segmentation algorithms.
413+ The last subfigures (F-G) show microscopic images from which we extracted the position of the
414+ bacterial agents and then made a prediction for the position of the rod-agents, which could
415+ now be compared with the experimental data (i.e. the next microscopic image) in order to estimate
416+ the parameters of the system.
417+
376418\begin {figure }[H]
377419 \centering
378420 \includegraphics [width=0.48\textwidth ]{figures/cr_mech_coli/bacterial-rods-0000000025.png}%
0 commit comments