Synchronized and fresh communication of common information is vitally important in numerous multi-user network scenarios, whereby end-users must perform coordinated real-time action with the available information. However, developing efficient policies with performance guarantees is greatly complicated by the abruptly changing nature of related age and synchronization metrics. In particular, powerful approaches that are based on so-called drift-plus-penalty (DPP) methods could not be employed due to the non-traditional multiplicative update dynamics of age and synchronization. In this paper, we overcome these limitations by designing and analyzing a Lyapunov-drift-based algorithm under the non-traditional age and synchronization dynamics that is not only low-complexity and analyzable, but also performs better than all the prior designs in numerical investigations. By comparing our design with two alternatives using the DPP approach, we also shed some light on the key aspect of our design that enables the performance analysis, which may be useful in future studies in multiplicative update dynamics. Furthermore, we implement Feature-based Reinforcement Learning (RL) methods with reduced state spaces and RL with full-state observations. Fresh-Async performs very closely to feature-based RL method using a feature space consisting of average age, age asynchrony, and maximum age, and both algorithms exhibit competitive performance compared to full-state RL while maintaining superior computational efficiency. These investigations clarify the contrast between drift-based and RL-based designs, and also reveal how drift-based design can be beneficial for feature selection for RL operation.

We study the optimal scheduling problem where n source nodes attempt to transmit updates over L shared wireless on/off fading channels to optimize their age performance under energy and age-violation tolerance constraints. Specifically, we provide a generic formulation of age-optimization in the form of a constrained Markov Decision Process (CMDP), and obtain the optimal scheduler as the solution of an associated Linear Programming problem. We investigate the characteristics of the optimal single-user multi-channel scheduler under different age-related objectives where a usual threshold-based policy does not apply. We then investigate the stability region of the optimal scheduler for the multi-user case under age-violation tolerance constraints. Furthermore, we develop two online schedulers that do not require statistics and are amenable to scalable operation: Drift-plus-penalty-based design, and a novel variation of the well-known Q-learning-based reinforcement learning method that combines Q-learning with drift-minimization-methods successfully for the first time, to the best of our knowledge. Our numerical studies compare the performance of our online schedulers to the optimal scheduler to reveal that both algorithms capture the essential behavior of the optimal design under different scenarios with good scalability, with the Q-learning-based design providing even closer performance to the optimal one by utilizing the history of the drift in a novel way.

We consider a scenario whereby the state of a common source is being updated at multiple distributed devices. We are particularly interested in the tradeoff that exists between thefreshnessof the updates at the distributed devices and thesynchronyof the updates across them. In this paper, we explore this tradeoff in a wireless downlink setting whereby the transmitter can choose between unicast transmissions (with given success probabilities) to particular users and broadcast transmissions (with a smaller success probability) to all users. After discussing the Linear Programming (LP)-based optimal design and extreme choices of “always-unicasting” and “always-broadcasting” policies, we note that the optimal design is not scalable and the extreme policies are inefficient. This motivates us to develop two classes of policies, namely a “mixed randomized policy” and a “feature-based learning policy”, which have desirable performance and computational-complexity characteristics. Additionally we manage to provide complete analysis for the mixed randomized policy under the two-user case, which provides interesting insights and can be partially extended to general cases. We perform extensive numerical studies to compare the performance of these designs over the benchmarks to reveal their gains.

We consider a scenario whereby the state of a common source is being updated at multiple distributed devices. We are particularly interested in the tradeoff that exists between the freshness of the updates at the distributed devices and the synchrony of the updates across them. In this paper, we explore this tradeoff in a wireless downlink setting whereby the transmit-ter can choose between unicast transmissions (with given success probabilities) to particular users and broadcast transmissions (with a smaller success probability) to all users. After discussing the Linear Programming (LP)-based optimal design and extreme choices of “always-unicasting” and “always-broadcasting” poli-cies, we note that the optimal design is not scalable and the extreme policies are inefficient. This motivates us to develop two classes of policies, namely a “mixed randomized policy” and a “feature-based learning policy”, which have desirable performance and computational-complexity characteristics. We perform extensive numerical studies to compare the performance of these designs over the benchmarks to reveal their gains.

Future generation wireless technologies are expected to serve an increasingly dense and dynamic population of users that generate short bundles of information to be transferred over the shared spectrum. This calls for new distributed and low-overhead Multiple-Access-Control (MAC) strategies to serve such dynamic demands with spectral efficiency characteristics. In this work, we address this need by identifying and developing two fundamentally different MAC paradigms: (i) congestion-based paradigm that estimates the congestion level in the system and adapts to it; and (ii) age-based paradigm that prioritizes demands based on their ages. Despite their apparent differences, we develop policies under each paradigm in a generic multi-channel access scenario that are provably throughput-optimal when they employ any asymptotically-efficient channel encoding/decoding mechanism. We also characterize the stability regions of the two designs, and investigate the conditions under which one design outperforms the other. We perform extensive simulations to validate the theoretical claims and investigate the non-asymptotic performances of our designs.

With the adoption of 5G wireless technology and the Internet-of-Things (IoT) networking, there is a growing interest in serving a dense population of low-complexity devices over shared wireless uplink channels. Different from the traditional scenario of persistent users, in these new networks each user is expected to generate only small bundles of information intermittently. The highly dynamic nature of such demand and the typically low-complexity nature of the user devices calls for a new MAC paradigm that is geared for low-overhead and distributed operation of dynamic users.In this work, we address this need by developing a generic MAC mechanism for estimating the number and coordinating the activation of dynamic users for efficient utilization of the time-frequency resources with minimal public feedback from the common receiver. We fully characterize the throughput and delay performance of our design under a basic threshold-based multi-channel capacity condition, which allows for the use of different channel utilization schemes. Moreover, we consider the Successive-Interference-Cancellation (SIC) Multi-Channel MAC scheme as a specific choice in order to demonstrate the performance of our design for a spectrally-efficient (albeit idealized) scheme. Under the SIC encoding/decoding scheme, we prove that our low-overhead distributed MAC can support maximum throughput, which establishes the efficiency of our design. Under SIC, we also demonstrate how the basic threshold-based success model can be relaxed to be adapted to the performance of a non-ideal success model.

Terrain-based localization is an alternate to the global positioning system (GPS) in signal blocked areas. However, terrain-based localization technique may suffer from low accuracy or even fail when brake vibration occurs. This paper presents a real-time algorithm for vehicle localization which is robust against brake vibration. The input includes a reference map of pitch difference and measurements from rear wheel encoders and inertial measurement units (IMU). This method consists of two steps. In the first step, terrain map is generated using pitch difference at equidistant intervals. After that, the Bayesian inference and particle filters are adopted in the second step to identify the vehicle location during travel. To enhance system stability, we propose dynamic distributions of filter variances according to acceleration input. Experimental results demonstrate that the localization method with dynamic distributions can localize the vehicle quickly with high accuracy even when a quite severe shuddering happens.