The paper is devoted to an investigation of properties of dynamic modal logics with some interesting and rich features. A dynamic modal logic DML and its extension \(\mu\)DML are presented. They were developed in order to unify and extend dynamic epistemic logics proposed before in another paper by the author. An alternative to Kripke models, called modal structures (proposed by Fagin and Vardi), is used in the paper as a semantics for DML. The combination of ``a simple yet expressive logic with a natural semantic concept'' is considered as a useful instrument for the analysis of the epistemic dynamics of multiagent systems. Some approaches to dynamization of modal (epistemic) logics are surveyed (Plaza, Gerbrandy and Groeneveld, Baltag, Van Ditmarsch). DML and \(\mu\)DML are aiming to unify the approaches mentioned in the survey. Moreover, DML and \(\mu\)DML also provide an extended expressive power: nonconservative actions (f.ex. reset) are introduced. Action modalities and ordinary modalities are unified in DML. The semantics is formulated both in Kripke models and modal structures, MS. A simplification of MS, lean modal structure, is introduced in the paper. Then a multimodal logic ML is described. The language of ML contains modalities \(a \in A\). Modal formulas are of the form \([a]\phi\), where \(\phi\) is a formula. A usual axiomatization of ML is given (all instances of propositional tautologies, modus ponens, necessitation and distribution of modalities over implication \([a](\phi \rightarrow \psi) \rightarrow ([a]\phi \rightarrow [a]\psi)\)). An interpretation of ML in MS is given and fundamental results (soundness, completeness and decidability) are summarized. Compactness fails for MS, but it is regained in \(\text{MS}_\omega\), a completion of MS. Dynamic modal logic (DML) extends ML with actions and action modalities. The interpretation of ML in MS is extended -- actions are interpreted as relations on MS, intuitively, an action \(\beta\) is a transition from \(f \in \text{MS}\) to \(g \in \text{MS}\). Interpretation of formulas is modified accordingly. An axiomatization of DML and its semantics in terms of Kripke models is given. Both approaches to semantics of DML (MS and Kripke models) are discussed and compared. \(\mu\)DML is an extension of DML with recursively defined actions. Syntax is extended by action variables and a new action constructor \(\mu x. \beta\) is introduced. A semantics in terms of \(\text{MS}_\omega\) is presented (a fixed-point construction is used for the interpretation of \(\mu x. \beta\)). The semantics of \(\mu\)DML can be given also in terms of Kripke models. Some interesting features of \(\mu\)DML are studied. It is shown how introspective update operator and learning operator are definable in \(\mu\)DML. It is shown that in certain cases the semantics of recursively defined actions can be given in (finite) MS. An extension of \(\mu\)DML with simultaneous modifications and a representation of action structures via simultaneous fixed point definitions are presented. \(\mu\)DML can be reduced to DML by eliminating fixed point definitions. It is shown that axiom schemata \([a]\phi \rightarrow \phi\), \([a]\phi \rightarrow [a][a]\phi\), \(\neg[a]\phi \rightarrow [a]\neg[a]\phi\) and \(\langle\rangle\top\) are refuted in MS. A partial answer to the general question ``what is the relation between axiom schemata and the actions preserving them?'' is given. The paper is concluded by some open issues.