Longitudinal studies are powerful tools to achieve a better understanding of temporal progressions of biological or natural phenomenons. For instance, efforts in chemotherapy monitoring rely more and more on the understanding of the global disease progression and not only on punctual states of health. Mixed effects models have proved their efficiency in the study of longitudinal data sets, especially for medical purposes. This talks presents a nonlinear mixed effects model which allows to estimate both a group-representative piecewise-geodesic trajectory and inter-individual variability. This model provides a generic and coherent framework for studying longitudinal manifold-valued data. Estimation is formulated as a well-defined and consistent Maximum A Posteriori (MAP). Numerically, due to the non-linearity of the proposed model, the MAP estimation of the parameters is performed through a stochastic version of the Expectation-Maximization algorithm. I will present a new version of the Stochastic-Approximation EM (SAEM) algorithm which prevent convergence toward local minima.