In recent years, by theory and observation cosmology has advanced substantially. Parameters of the concordance or $\Lambda$CDM cosmological model are given with unprecedented precision ("precision cosmology"). On the other side, 95% of the matter content of the universe are of an unknown nature. This awkward situation motivates the present attempt to find cosmology's place among the (exact) natural sciences. Due to its epistemic and methodical particularities, e.g., as a mathematized historical science, cosmology occupies a very special place. After going through some of the highlights of cosmological modeling, the conclusion is reached that knowledge provided by cosmological modeling cannot be as explicative and secure as knowledge gained by laboratory physics.

It is shown that the problem of a possible violation of the Lorentz transformations at Lorentz factors $\gamma >5\times 10^{10} ,$ indicated by the situation which has developed in the physics of ultra-high energy cosmic rays (the absence of the GZK cutoff), has a nontrivial solution. Its essence consists in the discovery of the so-called generalized Lorentz transformations which seem to correctly link the inertial reference frames at any values of $\gamma .$ Like the usual Lorentz transformations, the generalized ones are linear, possess group properties and lead to the Einstein law of addition of 3-velocities. However, their geometric meaning turns out to be different: they serve as relativistic symmetry transformations of a flat anisotropic Finslerian event space rather than of Minkowski space. Consideration is given to two types of Finsler spaces which generalize locally isotropic Riemannian space-time of relativity theory, e. g. Finsler spaces with a partially and entirely broken local 3D isotropy. The investigation advances arguments for the corresponding generalization of the theory of fundamental interactions and for a specific search for physical effects due to local anisotropy of space-time.

It is shown that space-time may be not only in a state which is described by Riemann geometry but also in states which are described by Finsler geometry. Transitions between various metric states of space-time have the meaning of phase transitions in its geometric structure. These transitions together with the evolution of each of the possible metric states make up the general picture of space-time manifold dynamics.