Multiuser Scheduling for Minimizing Age of Information in Uplink MIMO Systems

He Chen,Qian Wang,Zheng Dong,Ning Zhang

DOI: https://doi.org/10.48550/arXiv.2002.00403

2020-02-12

Abstract:This paper studies the user scheduling problem in a multiuser multiple-input multi-output (MIMO) status update system, in which multiple single-antenna devices aim to send their latest statuses to a multiple-antenna information-fusion access point (AP) via a shared wireless channel. The information freshness in the considered system is quantified by a recently proposed metric, termed age of information (AoI). Thanks to the extra spatial degrees-of-freedom brought about by the multiple antennas at the AP, multiple devices can be granted to transmit simultaneously in each time slot. We aim to seek the optimal scheduling policy that can minimize the network-wide AoI by optimally deciding which device or group of devices to be scheduled for transmission in each slot given the instantaneous AoI values of all devices at the beginning of the slot. To that end, we formulate the multiuser scheduling problem as a Markov decision process (MDP). We attain the optimal policy by resolving the formulated MDP problem and develop a low-complexity sub-optimal policy. Simulation results show that the proposed optimal and sub-optimal policies significantly outperform the state-of-the-art benchmark schemes.

Information Theory,Networking and Internet Architecture,Signal Processing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to minimize the timeliness metric of information - namely, the Age of Information (AoI) - in a multi - user, multi - input, multi - output (MIMO) state - update system. Specifically, the paper focuses on how to formulate an optimal user - scheduling strategy to minimize the long - term average AoI of the entire network when multiple single - antenna devices send the latest state information to an information - fusion access point (AP) equipped with multiple antennas through a shared wireless channel.

### Background and Motivation
- **AoI Concept**: AoI is an indicator for measuring the freshness of information, defined as the time interval from the generation time of the last successfully received state - update packet at the destination to the present. Different from traditional packet - based performance metrics such as delay and throughput, AoI can continuously capture the degree of information staleness from the perspective of the destination.
- **Research Status**: Early research on AoI mainly focused on using queuing theory to analyze the average AoI performance of point - to - point systems. Recent research has turned to the problem of minimizing the network - wide AoI in more practical multi - source networks.

### Research Problems
- **Multi - user Scheduling Problem**: The paper specifically focuses on how to decide which device or group of devices should be scheduled for transmission in each time slot in a multi - user MIMO state - update system, so as to minimize the AoI within the network range given the instantaneous AoI values of all devices at the start of the current time slot.
- **Technical Challenges**: In a multi - user MIMO system, scheduling more users to transmit in the same time slot will lead to an increase in the transmission error probability of each user. This is because the additional spatial degrees of freedom provided by multiple antennas can only be used to support more users or improve the transmission reliability of the scheduled users.

### Solutions
- **Markov Decision Process (MDP)**: The paper models the multi - user scheduling problem as an MDP problem and obtains the optimal scheduling strategy by solving this MDP problem.
- **Low - complexity Sub - optimal Strategy**: To reduce the computational complexity, the paper also proposes a low - complexity sub - optimal strategy.

### Experimental Results
- **Simulation Results**: The simulation results show that the proposed optimal and sub - optimal strategies are significantly better than the existing benchmark schemes, especially under high signal - to - noise ratio (SNR) conditions.

### Conclusions
- **Main Contributions**: The paper is the first to attempt to use multi - antenna technology to reduce AoI in multi - user networks and finds the optimal scheduling strategy through the MDP framework. In addition, a low - complexity sub - optimal strategy is proposed, whose performance is close to that of the optimal strategy.

### Formulas
- **Received Signal Model**:
\[
y=\sum_{i = 1}^{K}\sqrt{P_i}h_ix_i + n=Hx + n
\]
where \(H = [\sqrt{P_1}h_1,\ldots,\sqrt{P_K}h_K]\), \(x = [x_1,\ldots,x_K]^T\), \(n\sim\mathcal{CN}(0,\sigma^2I_N)\).

- **Zero - Forcing Receiver**:
\[
\hat{x}=H^{\dagger}y=x + H^{\dagger}n
\]
where \(H^{\dagger}=(H^HH)^{-1}H^H\) is the pseudo - inverse of \(H\).

- **Transmission Error Probability**:
\[
P_e(k)=1-\sum_{i = 0}^{N - k}\frac{\left(\frac{\sigma^2}{P\Omega\gamma_{\text{th}}}\right)^i}{i!}\exp\left(-\frac{\sigma^2}{P\Omega\gamma_{\text{th}}}\right)
\]

- **System Expected AoI**:
\[
\bar{\delta}_s=\lim_{T\rightarrow\infty}\frac{1}{KT}\sum_{t = 1}^{T}\sum_{i =