The use of effective power management strategies is essential in improving the power utilization efficiency and battery endurance of embedded systems. Accordingly, this paper presents a dynamic power management mechanism based on a reinforcement learning agent to adaptively manage the power consumption and service achievability of an embedded system device. The simulation results show that for a given set of environmental conditions, the proposed mechanism yields a notable improvement in both the power utilization efficiency and the battery endurance of the embedded system compared to that obtained using a static power management scheme.