Development of a predictive adaptive resource reallocation method with critical process dispatching in information systems on mobile platforms

This work investigates the process of real-time resource management of an information system on a mobile platform under intermittent connectivity, destructive influences, and nonstationary resource availability. The scientific task relates to the fact that forecast errors and the inertia of platform reconfiguration can induce oscillatory resource redistribution, causing critical processes to intermittently lose the minimum required resource at each control time-step. A time-step-based predictive adaptive redistribution method with dynamic reservation has been devised in this study. The method introduces an operational survivability constraint. It is formulated as a requirement to maintain a minimum guaranteed resource profile for critical processes. The method adapts the reservation volume according to the assessed reliability of the current forecast. It also constrains the frequency of reconfigurations within a sliding window. These constraints are combined with dispatching of critical processes by urgency and allowable delay. By combining forecast-adaptive reservation with inertia-consistent reconfiguration constraints, the control loop reduces the amplitude and the cumulative intensity of reconfigurations. In particular, under bursty critical workload, the maximum reconfiguration step decreases by about 52%, while the cumulative magnitude of profile changes decreases by about 14%. These effects are explained by the fact that reserved resources compensate for forecast degradation, whereas reconfiguration constraints prevent abrupt control actions and stabilize time-step allocation. The results could be implemented to build resource management information systems for robotic platforms, sensor networks, as well as mobile systems under intermittent connectivity and real-time resource degradation.