“…As is well known, risk-sensitive control is a subject of significant interest in the field of Markov decision processes (MDPs) and has received much attention; see, for example, [2], [5], [8], and [14] for average cost discrete-time MDPs (DTMDPs), [2], [7], and [13] for infinitehorizon discounted cost DTMDPs, [2] and [7] for finite-horizon discounted or undiscounted cost DTMDPs, [10] and [22] for average cost continuous-time MDPs (CTMDPs), and [10] for finite-horizon cost and infinite-horizon discounted cost CTMDPs. Regarding risk-sensitive semi-Markov decision processes (SMDPs), to the best of the authors' knowledge this issue was only addressed in [6], in which the authors concentrated on a long-run average cost. In this paper we also study risk-sensitive SMDPs, but we focus on a finite-horizon total cost.…”