A UoI-Optimal Policy for Timely Status Updates with Resource Constraint.

Area of Science:

• Computer Science
• Control Systems Engineering
• Information Theory

Background:

• Timely status updates are crucial for remote control systems like autonomous driving and the Industrial Internet of Things (IIoT).
• Existing Age of Information (AoI) metrics do not fully capture context-dependent timeliness requirements, especially during emergencies.
• The Urgency of Information (UoI) metric was introduced to incorporate context-aware weights, reflecting process criticality.

Purpose of the Study:

• To develop an optimal updating and scheduling policy based on the Urgency of Information (UoI) metric under resource constraints.
• To address the open problem of UoI-optimal strategies in dynamic remote control systems.
• To compare the proposed UoI-optimal policy against existing AoI-optimal policies.

Main Methods:

• Formulated the problem as a constrained Markov decision process to derive a UoI-optimal policy with a threshold structure.
• Developed a numerical method using linear programming for scenarios with known context-aware weights.
• Designed a reinforcement learning (RL)-based scheduling policy for unknown system models and weights.

Main Results:

• The UoI-optimal policy exhibits a threshold structure, with the threshold increasing as resource constraints tighten.
• The proposed UoI-optimal policy significantly reduces the average squared estimation error compared to AoI-optimal policies.
• The RL-based updating policy demonstrates near-optimal performance without requiring prior knowledge of the system model.

Conclusions:

• The UoI metric provides a more effective framework for optimizing data updates in time-sensitive remote systems.
• The developed threshold-based policy and RL approach offer practical solutions for resource-constrained, context-aware information updating.
• The research advances the state-of-the-art in real-time data management for critical applications like autonomous driving and IIoT.