A dynamic Bayesian network is a Bayesian network that overcomes the inability to model feedback loops.  In this case, the variables are replicated for each time step and the feedback is modeled by connecting the nodes at adjacent time steps.