Abstract: Understanding how cells position themselves correctly in multicellular environments is a fundamental challenge in developmental biology. We use the invariable embryogenesis of C. elegans and a modelling approach to better understand the underlying mechanisms driving cellular movements. We developed a simulator to model the physical interactions among cells that influence cell motion, including physical constraints, cell division, adhesion, and active forces. To inform and evaluate simulations we use a dataset of real embryos with microscopy tracked nuclei positions. In the model cells are represented as soft spheres that can repulse and adhere to each other, and also move and divide. Cell division timings and movement directions are taken from the dataset. The egg shell repulses the cells and is modelled as a convex hull whose shape is derived from the embryo dataset. We use the model to unravel the contribution of purely physical processes in early embryogenesis, such as the spatial constraints, egg shape, and shape changes due to divisions, and active biological phenomena such as differential adhesion and active movements. The problem is framed as a machine learning approach to find the correct cellular positioning, where hypotheses are evaluated by learning optimal parameter values on a training set and testing how much they improved cellular positioning. To optimize the model parameters we used evolutionary algorithms. Differential evolution is used for continuous variables and a novel molecule-based evolution scheme optimizes the differential cellular adhesion. At the 8-cell stage our simulations highlight previously described phenomena, such as a weak basal adhesion force among all cells and active movements for the ABpl and ABpr cells. Also adhesion between E and P3 is suggested, which matches published results of how descendants of both cells adhere during gastrulation. Our main goal however is to reproduce the complex movements during the AB64 stage. We are currently evaluating our predictions for this phase with experimens, by visualizing e-cadherin molecules, cell shapes and the actomyosin network. As a perspective, we are expanding our simulator with a signaling model. Cell signaling is represented as a network of reactions expressed as logic rules, where signals trigger cascades and determine cell fate. By modelling how signals affect fate, linking fate to cell behavior, we can predict the effects of perturbations on cell fates and positions.