[Paper Review] Optimizing Data Freshness in Time-Varying Wireless Networks with Imperfect Channel State

Haoyue Tang, Jintao Wang+2 more|arXiv (Cornell University)|Oct 6, 2019

Age of Information Optimization11 references4 citations

TL;DR

This paper proposes an asymptotically optimal scheduling algorithm for minimizing Age of Information (AoI) in time-varying, bandwidth-limited wireless networks with imperfect channel state estimation and packet loss. By modeling the problem as a constrained Markov decision process and using linear programming to optimize single-user transmission strategies, the method dynamically schedules sensors based on channel quality and AoI, outperforming greedy policies and showing that sensors with poor channels are prioritized at higher AoI to reduce loss probability.

ABSTRACT

We consider a scenario where a base station (BS) attempts to collect fresh information from power constrained sensors over time-varying band-limited wireless channels. We characterize the data freshness through the recently proposed metric--the Age of Information. We consider a time-varying channel model with power adaptation. Unlike previous work, packet loss may happen due to imperfect channel estimation or decoding error. We propose an asymptotic optimal scheduling algorithm minimizing AoI performance and satisfying both bandwidth and power constraint in such networks. Numerical simulations show that the proposed policy outperforms the greedy one, and we observe that sensors with poor channels are scheduled at higher AoI in order to limit the packet loss.

Motivation & Objective

• To address the challenge of maintaining data freshness in time-varying, bandwidth-constrained wireless networks with imperfect channel state information and non-zero packet loss.
• To design a cross-layer scheduling strategy that minimizes the Age of Information (AoI) under joint bandwidth and average power constraints.
• To account for realistic transmission errors due to imperfect channel estimation or decoding, which are often ignored in prior AoI-optimization works.
• To develop a practical multi-user scheduling policy that is asymptotically optimal as the number of sensors increases.

Proposed method

• Decomposes the multi-user scheduling problem into single-user constrained Markov decision processes (CMDPs) to simplify optimization.
• Models transmission success probability as a function of estimated channel state, incorporating packet loss due to imperfect CSI or decoding errors.
• Uses linear programming (LP) to solve the single-user CMDP and derive the optimal transmission policy for each sensor.
• Constructs a relaxed scheduling policy (π_R) that satisfies bandwidth and power constraints, enabling theoretical performance analysis.
• Proposes a truncated scheduling policy (π̂) that respects the bandwidth limit M by randomly dropping scheduled sensors when more than M are selected.
• Analyzes the performance gap between the truncated policy and the relaxed optimal policy using probabilistic bounds on AoI and scheduling deviation.

Experimental results

Research questions

• RQ1 How can AoI be minimized in time-varying wireless networks when channel state information is imperfect and packet loss occurs?
• RQ2 What is the optimal trade-off between scheduling frequency and channel quality when transmission reliability is limited by estimation error or fading?
• RQ3 Can a scheduling policy be designed that is asymptotically optimal in large-scale networks with bandwidth and power constraints?
• RQ4 How does the inclusion of packet loss affect the optimal scheduling threshold and transmission strategy compared to error-free models?

Key findings

• The proposed scheduling policy outperforms greedy strategies in numerical simulations, especially in high-fading or high-loss environments.
• Sensors with poor channel states are scheduled at higher AoI to reduce the probability of packet loss, which improves overall freshness.
• The performance gap between the proposed policy and the optimal relaxed policy scales as O(1/√N), proving asymptotic optimality as the number of sensors increases.
• The method achieves a bounded performance loss even when the bandwidth constraint is tight, due to careful handling of scheduling deviation and AoI distribution.
• The analysis shows that the expected AoI deviation from the optimal policy diminishes with network size, validating the scalability of the approach.