Predictive Goal-Oriented Communication Policy for Remote Tracking of Markovian Sources

We investigate the remote tracking of a Markovian source in an energy-harvesting wireless system within the goal-oriented communication framework. We propose a Predictive transmission policy that exploits the source dynamics to suppress redundant updates while preserving energy for correcting persistent errors. The proposed policy employs prediction at the transmitter to anticipate the receiver's self-synchronization, thereby avoiding unnecessary transmissions. Simulations across different source dynamics and channel conditions show that, under severe energy constraints, the Predictive policy outperforms a reactive baseline that transmits whenever source and receiver states diverge, achieving low reconstruction error with reduced energy consumption.