Discounted continuous-time Markov decision process with constraints: unbounded transition and loss rate
Seminar Room 1, Newton Institute
It is known that the convex analytic approach is effective for solving constrained problems for discrete-time MDP. If the transition rate of a continuous-time model is bounded, then one can use the uniformization technique, but if that is unbounded, one has to investigate the occupation measures and linear programs from scratch. In this report, we present the main properties of the space of occupation measures, study the linear program thereon, and consider several examples.