The cosmological observations suggest that the presently accelerating universe should be filled by an exotic form of matter, violating the strong energy condition, of unknown nature and origin. We propose the viscous dark matter of a source of acceleration in the form of Chaplygin gas which is characterized by equation of state in the phenomenological form $p=-\frac{A}{\rho^{\alpha}}$, where $p$ and $\rho$ are pressure and energy density respectively ($A$ and $\alpha$ are constants). Chaplygin gas is interpreted in terms of viscous matter and without the cosmological constant. The acceleration effect is caused only by viscosity in this class of cosmological models. We show that bulk viscosity effects introduced to the standard FRW cosmology give rise to the natural unification of both dark matter and dark energy. We show that dust viscous cosmological models are structurally stable if $m < 1/2$ ($1+\alpha=1/2-m$).

Recent astronomical observations have indicated that the Universe is in the phase of accelerated expansion. While there are many cosmological models which try to explain this phenomenon, we focus on the interacting $\Lambda$CDM model where the interaction between the dark energy and dark matter sectors takes place. This model is compared to its simpler alternative---the $\Lambda$CDM model. To choose between these models the likelihood ratio test was applied as well as the model comparison methods (employing Occam's principle): the Akaike information criterion (AIC), the Bayesian information criterion (BIC) and the Bayesian evidence. Using the current astronomical data: SNIa (Union2.1), $h(z)$, BAO, Alcock--Paczynski test and CMB we evaluated both models. The analyses based on the AIC indicated that there is less support for the interacting $\Lambda$CDM model when compared to the $\Lambda$CDM model, while those based on the BIC indicated that there is the strong evidence against it in favor the $\Lambda$CDM model. Given the weak or almost none support for the interacting $\Lambda$CDM model and bearing in mind Occam's razor we are inclined to reject this model.

We study the dynamical evolution of cosmological models with the Robertson-Walker symmetry with a scalar field non-minimally coupled to gravity and barotropic matter. For this aim we use dynamical system methods. We have found a type of evolutional path which links between all important events during the evolution, the cosmological singularity of finite time, inflation, radiation and matter dominating epoch and the accelerated phase expansion of the universe. We point out importance of finding the new generic solution called a twister solution for a deeper description of the evolution of the Universe. We demonstrate that including the non-minimal coupling leads to a new, richer evolutional cosmological scenario in comparison to the case of minimal coupling.

We study the general chaotic features of dynamics of the phantom field modelled in terms of a single scalar field conformally coupled to gravity. We demonstrate that the dynamics of the FRW model with dark energy in the form of phantom field can be regarded as a scattering process of two types: multiple chaotic and classical non-chaotic. It depends whether the spontaneously symmetry breaking takes place. In the first class of models with the spontaneously symmetry breaking the dynamics is similar to the Yang-Mills theory. We find the evidence of a fractal structure in the phase space of initial conditions. We observe similarities to the phenomenon of a multiple scattering process around the origin. In turn the class of models without the spontaneously symmetry breaking can be described as the classical non-chaotic scattering process and the methods of symbolic dynamic are also used in this case. We show that the phantom cosmology can be treated as a simple model with scattering of trajectories which character depends crucially on a sign of a square of mass. We demonstrate that there is a possibility of chaotic behavior in the flat Universe with a conformally coupled phantom field in the system considered on non-zero energy level. We obtain that the acceleration is a generic feature in the considered model without the spontaneously symmetry breaking. We observe that the effective EOS coefficient oscillates and then approach to $w=-1$.

Recent astronomical observations indicate that the Universe is presently almost flat and undergoing a period of accelerated expansion. Basing on Einstein's general relativity all these observations can be explained by the hypothesis of a dark energy component in addition to cold dark matter (CDM). Because the nature of this dark energy is unknown, it was proposed some alternative scenario to explain the current accelerating Universe. The key point of this scenario is to modify the standard FRW equation instead of mysterious dark energy component. The standard approach to constrain model parameters, based on the likelihood method, gives a best-fit model and confidence ranges for those parameters. We always arbitrary choose the set of parameters which define a model which we compare with observational data. Because in the generic case, the introducing of new parameters improves a fit to the data set, there appears the problem of elimination of model parameters which can play an insufficient role. The Bayesian information criteria of model selection (the AIC and BIC) are dedicated to promotion a set of parameters which should be incorporated to the model. We divide class of all accelerating cosmological models into two groups according to the two types of explanation acceleration of the Universe. Then the Bayesian framework of model selection is used to determine the set of parameters which gives preferred fit to the SNIa data. We find a few of flat cosmological models which can be recommend by both the Bayes factor and Akaike information criterion.

We study the Friedmann-Robertson-Walker model with phantom fields modelled in terms of scalar fields. We apply the Ziglin theory of integrability and find that the flat model is nonintegrable. Then we cannot expect to determine simple analytical solutions of the Einstein equations. We demonstrate that there is only a discrete set of parameters where this model is integrable. For comparison we describe the phantoms fields in terms of the barotropic equation of state. It is shown that in the contrast to the phantoms modelled as scalar fields, the dynamics is always integrable and phase portraits are constructed. In this case we find the duality relation.

We study metric-affine gravity (MAG) inspired cosmological models. Those models were statistically estimated using the SNIa data. We also use the cosmic microwave background observations and the big-bang nucleosynthesis analysis to constrain the density parameter $\Omega_{\psi,0}$ which is related to the non-Riemannian structure of the underlying spacetime. We argue that while the models are statistically admissible from the SNIa analysis, complementary stricter limits obtained from the CMB and BBN indicate that the models with density parameters with a $a^{-6}$ scaling behaviour are virtually ruled out. If we assume the validity of the particular MAG based cosmological model throughout all stages of the Universe, the parameter estimates from the CMB and BBN put a stronger limit, in comparison to the SNIa data, on the presence of non-Riemannian structures at low redshifts.

In the present paper we raise the question whether initial cosmological singularity can be proved from the cosmological tests. The classical general relativity predict the existence of singularity in the past if only some energy conditions are satisfied. On the other hand the latest quantum gravity applications to cosmology suggest of possibility of avoiding the singularity and replace it with the bounce. The distant type Ia supernovae data are used to constraints on bouncing evolutional scenario where square of the Hubble function $H^2$ is given by formulae $H^2=H^2_0[\Omega_{m,0}(1+z)^{m}-\Omega_{n,0}(1+z)^{n}]$, where $\Omega_{m,0}, \Omega_{n,0}>0$ are density parameters and $n>m>0$. We show that the on the base of the SNIa data standard bouncing models can be ruled out on the $4\sigma$ confidence level. If we add the cosmological constant to the standard bouncing model then we obtain as the best-fit that the parameter $\Omega_{n,0}$ is equal zero which means that the SNIa data do not support the bouncing term in the model. The bounce term is statistically insignificant the present epoch. We also demonstrate that BBN offer the possibility of obtaining stringent constraints of the extra term $\Omega_{n,0}$. The other observational test methods like CMB and the age of oldest objects in the Universe are used. We also use the Akaike informative criterion to select a model according to the goodness of fit and we conclude that this term should be ruled out by Occam's razor, which makes that the big bang is favored rather then bouncing scenario.

We consider a Friedmann-Robertson-Walker cosmological model with some exotic perfect fluid with spin known as the Weyssenhoff fluid. The possibility that the dark energy may be described in part by the Weyssenhoff fluid is discussed. The observational constraint coming from supernovae type Ia observations is established. This result indicates that, whereas the cosmological constant is still needed to explain current observations, the model with spin fluid is admissible. For high redshifts $z > 1$ the differences between the model with spin fluid and the cold dark matter model with a cosmological constant become detectable observationally for the flat case with $\Omega_{\text{m},0}=0.3$. From the maximum likelihood method we obtain the value of $\Omega_{\text{s},0} = 0.004 \pm 0.016$. This gives us the limit $\Omega_{\text{s},0} > -0.012$ at the $1\sigma$ level. While the model with ``brane effects'' is preferred by the supernovae Ia data, the model with spin fluid is statistically admissible. For comparison, the limit on the spin fluid coming from cosmic microwave background anisotropies is also obtained. The uncertainties in the location of a first peak give the interval $-1.4 \times 10^{-10} < \Omega_{\text{s},0} < -10^{-10}$. From big bang nucleosynthesis we obtain the strongest limit $\Omega_{\text{s},0} \gtrsim -10^{-20}$. The interconnection between the model considered and brane models is also pointed out.

We show the effectiveness of using the procedure of the potential reconstruction (the V-method) in the estimation of the transition redshift corresponding to switch from deceleration to acceleration phase of the Universe. We investigate the FRW models with dark energy by applying the particle-like description of their dynamics. In this picture the evolution of the universe is represented by motion of a fictitious particle in the one-dimensional potential $V(a)$ or V(z)$ where $a$ is the scale factor and $z$ is the redshift. The V-method solves the inverse problem where we reconstruct the potential function from empirical data. We use the Riess et al.'s gold and gold+silver samples of SN Ia data in this reconstruction. In the same framework we obtain both the estimates of the form of the equation of state for dark energy and the transition redshift. We obtain that transition redshift is $0.38_{-0.07}^{+0.10}$ (gold subset) and $0.37_{-0.07}^{+0.06}$ (gold+silver sample) when the linear model of dark energy ($w(z)=w_{0} + w_{1}z$) and $\Omega_{m,0}=0.3$ are assumed. We compare the estimation of transition redshift with the Riess et al. results. The V-method also alows us to find the value of Hubble function at the moment of transition.

We discuss some observational constraints resulting only from SNIa observations, imposed on the behavior of the original flat Cardassian model, and its extension with the curvature term included. We test the models using the Perlmutter SN Ia data as well as the new Knop and Tonry samples. We estimate the Cardassian model parameters using the best-fitting procedure and the likelihood method. In the fitting procedure we use the density variables for matter, Cardassian fluid and curvature, and include the errors in redshift measurements. For the Perlmutter sample in the non-flat Cardassian model we obtain the high or normal matter density universe ($\Omega_{m,0} \ge 0.3$), while for the flat Cardassian model we have the high density universe. For sample A in the high density universe we also find the negative values of estimates of $n$ which can be interpreted as the phantom fluid effect. For the likelihood method we get that a nearly flat universe is preferred. We show that, if we assume that the matter density is 0.3, then $n \approx 0$ in the flat Cardassian model, which corresponds to the Perlmutter model with the cosmological constant. Testing with the Knop and Tonry SN Ia samples show no significant differences.

We apply methods of dynamical systems to study the behaviour of the Randall-Sundrum models. We determine evolutionary paths for all possible initial conditions in a 2-dimensional phase space and we investigate the set of accelerated models. The simplicity of our formulation in comparison to some earlier studies is expressed in the following: our dynamical system is a 2-dimensional Hamiltonian system, and what is more advantageous, it is free from the degeneracy of critical points so that the system is structurally stable. The phase plane analysis of Randall-Sundrum models with isotropic Friedmann geometry clearly shows that qualitatively we deal with the same types of evolution as in general relativity, although quantitatively there are important differences.