NextG applications such as AR/VR, industrial automation, cloud gaming, and autonomous robots increasingly demand lower latencies. Current 5G networks, however, incur significant delays due to request-based scheduling, where users must signal demand before the base station can allocate resources for uplink transmissions. In this paper, we present Sage, a real-time AI system that can predict per-user uplink demand at millisecond granularity and proactively allocate resources to reduce uplink latency. Sage proposes traffic trains: a novel abstraction that mitigates distortions to the observed traffic arrivals at the base station and yields stable prediction targets. Sage extracts statistical features from user traffic and retrieves appropriate models from a traffic-aware database of dedicated AI predictors. Sage further executes low-latency inference, error tracking, and online continual learning to dynamically adapt prediction models. Extensive evaluation shows that Sage achieves millisecond-level prediction accuracy with sub-millisecond inference overhead, reducing uplink latency by 2.53X on average across diverse applications while maintaining high resource efficiency.

We are witnessing a significant shift in social media platforms; we are transitioning from chronological social media feeds to feeds that are driven by AI recommendation systems. While the main goal of AI recommendation systems is to suggest engaging content to users, there are also some associated risks: AI recommendation systems can promote extreme content, causing negative consequences like online polarization and user radicalization. Overall, there is a pressing need to design powerful techniques that allow us to audit AI recommendation systems. Motivated by this, our work introduces ClipMind, a scalable and generalizable framework using advanced AI models to audit these recommendation algorithms on short-format video platforms like TikTok and YouTube Shorts. We demonstrate the merits of our framework by collecting social media feeds from TikTok. Our analysis shows that TikTok’s recommendation algorithm increasingly recommends similar videos when a user expresses interest in mainstream topics like Food and Beauty Care. On the other hand, by investigating niche interests (War and Mental Health), we find no evidence of informational rabbit holes of extreme content on TikTok. Our work contributes to efforts that leverage AI for social good, as our framework can be used by several interested stakeholders, including users, social media platforms, regulators, and researchers, to understand and audit video-based algorithmic recommendations.

This paper considers utilizing the knowledge of age gains to reduce the average age of information (AoI) in random access with event-driven periodic updating for the first time. Built on the form of slotted ALOHA, we require each device to determine its age gain threshold and transmission probability in an easily implementable decentralized manner, so that the contention can be limited to devices with age gains as high as possible. For the basic case that each device utilizes its knowledge of age gain of only itself, we provide an analytical modeling by a multi-layer discrete-time Markov chains (DTMCs), where an external DTMC manages the jumps between the beginnings of frames and an internal DTMC manages the evolution during an arbitrary frame, for obtaining optimal fixed access parameters offline. For the enhanced case that each device utilizes its knowledge of age gains of all the devices, we require each device to adjust its access parameters for maximizing the estimated network expected AoI reduction per slot, through maintaining the a posteriori joint probability distribution of local age and age gain of an arbitrary device in a Bayesian manner. Numerical results validate our study and demonstrate the advantage of the proposed schemes over other schemes.

Ultra-Reliable Low-Latency Communication (URLLC) was introduced in 5G to meet the demanding requirements of latencies as low as 0.5 ms and reliability of 99.999 % for specific applications. Despite over a decade of discussions on URLLC, achieving these standards in real-world implementations remains challenging. We argue that it is unclear if and how URLLC can be attained and a holistic system-level perspective that addresses all the system's inherent bottlenecks is needed. Inspired by a real-world 5G testbed, we present this comprehensive vision and show how to achieve latency requirements by outlining the necessary design choices across all system layers, including the processing and radio units.

5G enables diverse services through network slicing, allowing multiple virtual networks to share physical infrastructure. However, efficiently managing resources across slices is challenging. This demo presents SliceGuard, a two-level scheduling system that leverages WebAssembly (Wasm) to allow slice owners to run customized schedulers in a secure, platform-independent environment, while the network operator manages inter-slice resource allocation. We demonstrate dynamic slicing for cloud gaming over 5G and show how Wasm enables real-time scheduler updates and fault isolation. This approach enhances flexibility, security, and customization for private 5G networks.

Reconfigurable data center networks (RDCNs) are arising as a promising data center network (DCN) design in the post-Moore's law era. However, the constantly reconfigured network topology in RDCNs invalidates the assumption of using hop count as the cost metric for routing, e.g., the status quo Equal-Cost Multi-Path routing (ECMP) in traditional DCNs. Unfortunately, existing routing solutions in RDCNs stick to the old assumption and deliver suboptimal performance either high in latency or low in bandwidth efficiency. In this paper, we redefine the cost metric for RDCN routing with uniform cost to unify the effects of topology disruption and hop count on latency and bandwidth efficiency. We propose Uniform-Cost Multi-Path routing (UCMP), an ECMP equivalent for RDCNs, where minimizing uniform cost leads flows of various sizes to the right balance between latency and bandwidth efficiency. Our simulation shows that UCMP achieves 53% to 98% lower flow completion time (FCT) and 1.55× bandwidth efficiency compared to the state-of-the-art RDCN routing strategy, and our testbed implementation demonstrates sustainable switch resource usage of UCMP as RDCNs scale.

This paper considers designing an optimal policy for deadline-constrained access in cognitive radio networks, where a secondary user needs to complete a packet transmission over the vacant spectrum within a delivery deadline. To minimize the total access cost, it is desirable to design an optimal opportunistic access policy by utilizing channel dynamics and sensing outcomes. We take non-negligible switching overheads, a state-dependent overtime penalty, and practical switching operations into consideration in the Markov decision process formulation of such an access problem under wide-band sensing. Moreover, we establish the existence of monotone optimal decision rules to reduce the complexity of computing an optimal policy. Simulation results verify our theoretical studies and the cost advantage over other policies.

This paper considers a wireless uplink network consisting of multiple end devices and an access point (AP). Each device monitors a physical process with randomly generated status updates and sends these update packets to the AP in the uplink. The AP aims to schedule the transmissions of these devices to optimize the network-wide information freshness, quantified by the age of information (AoI) metric. Due to the stochastic arrival of the status updates at end devices, the AP only haspartial observationsof system times of the latest status update packets at end devices when making scheduling decisions. Such a decision-making problem can be naturally formulated as a partially observable Markov decision process (POMDP). We reformulate the POMDP into an equivalent belief Markov decision process (belief-MDP), by defining fully observable belief states of the POMDP as the states of the belief-MDP. The belief-MDP in its original form is difficult to solve as the dimension of its states can go to infinity and its belief space is uncountable. Fortunately, by carefully leveraging the properties of the status update arrival processes (i.e., Bernoulli processes), we manage to simplify the belief-MDP substantially, where every feasible state is characterized by a two-dimensional vector. Based on the simplified belief-MDP, we devise a low-complexity scheduling policy, termed Partially Observing Max-Weight (POMW) policy, for the formulated AoI-oriented scheduling problem. We derive upper bounds for the time-average AoI performance of the proposed POMW policy. We analyze the performance guarantee for the POMW policy by comparing its performance with a universal lower bound available in the literature. Numerical results validate our analyses and demonstrate that the performance gap between the POMW policy and its fully observable counterpart is proportional to the inverse of the lowest arrival rate of all end devices.

Designing efficient random access is a vital problem for urgency-constrained packet delivery in uplink Internet of Things (IoT), which has not been investigated in depth so far. In this paper, we focus on unpredictable frame-synchronized traffic, which captures a number of scenarios in IoT communications, and generalize prior studies on this issue by considering a general ALOHA-like protocol, a general single-packet reception (SPR) channel, urgency-dependent throughput (UDT) based on a general urgency function, and the dynamic programming optimality. With a complete knowledge of the number of active users, we use the theory of Markov Decision Process (MDP) to explicitly obtain optimal policies for maximizing the UDT, and prove that a myopic policy is in general optimal. With an incomplete knowledge of the number of active users, we use the theory of Partially Observable MDP (POMDP) to seek optimal policies, and show that a myopic policy is in general not optimal by presenting a counterexample. Because of the prohibitive complexity to obtain optimal or near-optimal policies for this case, we propose two practical policies that utilize the inherent property of our MDP framework and channel model. Simulation results show that both outperform other alternatives. The robustness under relaxed system settings is also examined.

In this paper, we consider a multiuser uplink status update system, where a monitor aims to timely collect randomly generated status updates from multiple end nodes through a shared wireless channel. We adopt the recently proposed metric, termed age of information (AoI), to quantify the information timeliness and freshness. Due to the random generation of the status updates at the end node side, the monitor only grasps a partial knowledge of the status update arrivals. Under such a practical scenario, we aim to address a fundamental multiuser scheduling problem: how to schedule the end nodes to minimize the network-wide AoI? To solve this problem, we formulate it as a partially observable Markov decision process (POMDP), and develop a dynamic programming (DP) algorithm to obtain the optimal scheduling policy. By noting that the optimal policy is computationally prohibitive, we further design a low-complexity myopic policy that only minimizes the one-step expected reward. Simulation results show that the performance of the myopic policy approaches that of the optimal policy and is superior to that of the baseline policy.

This paper considers a multiple-access system with multiple-packet reception (MPR) capability γ, i.e., a packet can be successfully received as long as it overlaps with γ -1 or fewer other packets at any instant during its lifetime. To efficiently utilize the MPR capability, this paper generalizes p-persistent carrier-sense multiple access (CSMA) to consider that a user with carrier sensing capability c adopts the transmission probability p, if this user has sensed n ongoing transmissions for n = 0, 1,⋯, c - 1. This paper aims to model the characteristics of such CSMA and to design transmission probabilities for achieving maximum saturation throughput. To this end, we first formulate such CSMA as a parameterized Markov decision process (MDP) and use the long-run average performance to evaluate the saturation throughput. Second, by observing that the exact values of optimal transmission probabilities are in general infeasible to find, we modify this MDP to establish an upper bound on the maximum throughput, and modify this MDP again to propose a heuristic design with near-optimal performance. Simulations with respect to a wide range of configurations are provided to validate our study. The throughput performance under more general models and the robustness of our design are also investigated.