Goal-oriented remote tracking with satellite communication

Abstract

The increasing deployment of sensors in remote areas has increased the importance of efficient tracking systems. Although existing studies often assume stationary monitors with fixed wireless communication channels, these assumptions are impractical in environments where wired communication cannot be implemented due to high infrastructure costs and environmental constraints, and where wireless communication is not feasible because the devices are located beyond the range of cellular networks. In these scenarios, satellite communications serve as a viable solution, with the satellite acting as an access point to establish a connectivity for remote monitoring. However, this introduces an additional complexity of the time-varying wireless channel conditions resulting from satellite motion. To overcome this challenge, this thesis research focuses on goal-oriented remote monitoring with satellite communications. In this study, a new system model is proposed and the system model consists of a source, a sensor, and a monitor, with the monitor mounted on a satellite. The sensor, powered by a battery recharged via an energy harvester, consists of a sampler, buffer, controller, and transmitter. The controller, without direct access to the source information, decides whether to sample the source or transmit the buffered data to the monitor. Due to satellite motion, the wireless channel conditions change over time. In mathematical modeling of system components, the source is modeled as a Markov chain with information states, and the satellite motion is represented as a Markov chain based on spatially dependent channel conditions. The overall system is formulated as a Partially Observable Markov Decision Process (POMDP), which is then transformed into a Markov Decision Process (MDP). The MDP is solved using the Relative Value Iteration (RVI) algorithm to derive the optimal control policy. This policy aims to optimize long-term energy consumption while minimizing the age of information at the monitor. Furthermore, system simulations are evaluated for three scenarios: asymmetric Markov sources, symmetric Markov sources, and a simulation scenario with two-state, four-state, and six-state Markov models for satellite motion. Furthermore, three estimation techniques, namely Maximum Likelihood Estimation (MLE), Minimum Mean Distortion Estimation (MMDE), and Minimum Mean Squared Error Estimation (MMSEE), are used by the monitor to estimate the information state at the source. The results show that the optimal policies derived for MLE, MMDE and MMSEE outperform baseline policies, in all the source scenarios. Moreover, simulation results suggest that the proposed system model achieves improved performance compared to the reference model, since the controller is capable of commanding increased data transmission during satellite motion states that are associated with higher Reception Success Rate (RSR).