Collaborative Research: CNS Core: Medium: Information Freshness in Scalable and Energy Constrained Machine to Machine Wireless Networks

 

Last updated in August 2024.

Project Team

·         Principal Investigators: Chih-Chun Wang (Purdue), Bo Ji (Virginia Tech), Ashutosh Sabharwal (Rice), and Ness Shroff (OSU).

·         This is a collaborative project among Purdue, OSU, Rice, and Virginia Tech, under the Grant Numbers CNS-2107363, CNS-2106932, CNS- 2106993 and CNS-2106427, respectively.

Vision

 

The last decade has seen an unprecedented growth in wireless devices and consumers' need to own and use them in new and exciting ways. As these devices become more popular, Machine Type Communications (MTC) are getting increasingly prevalent, where communications (through timely updates) occur primarily between machines. In such MTC systems, timely updates are needed to make decisions for various applications and actuations, and the utility of each update crucially depends on the freshness of information content. Furthermore, many of these IoT devices are battery powered, therefore mechanisms are needed in place to simultaneously ensure long battery lifetimes. Taking all of these considerations into account, the overarching goal of this project is to develop novel tools and techniques to optimize information freshness and energy efficiency while meeting the practical requirements of wireless MTC systems.

 

The goals of this project are to develop the analytical foundations for controlling and optimizing information freshness in wireless MTC networks, resulting in fully distributed provably efficient algorithms and protocols that will be extensively evaluated on a large-scale programmable testbed at Rice University.

 

By addressing the robustness, scalability, and energy efficiency in MTC networks, the results of this project will have significant impacts on a broad class of applications supported by these networks ranging from autonomous driving to smart healthcare. In addition to broadening its societal impact via analytical and empirical developments, a significant part of this project is to recruit women and under-represented minority students through research engagements and outreach activities within the four collaborating universities, including mechanisms to increase leadership and participation from under-represented groups.

 

 

Key Accomplishments

 

We conducted research on the following aspects of Age-of-Information (AoI) optimization in MTC networks: robustness, energy efficiency, and age-based scheduling.

 

Robustness in MTC Networks

 

·         We consider a status update system, in which update packets are sent to the destination via a wireless medium that allows for multiple rates, where a higher rate also naturally corresponds to a higher error probability. We design a low-complexity optimal scheduler that selects between two different transmission rate and error probability pairs to be used at each transmission epoch. To this end, we show that there exists a threshold-type policy that is age-optimal, and that the objective function is quasi-convex or non-decreasing in the threshold.

·         We consider a sampling problem, where the source generates fresh samples and sends the samples through a forward channel to a remote estimator, and acknowledgements are sent back over a feedback channel. Due to various channel conditions, the forward channel can be unreliable, and both the forward and feedback channels could have random transmission delays. This problem is motivated by distributed sensing, where the estimator can receive the signal samples from the channel and the noisy observation from a local sensor. It is shown that the estimation error is a non-decreasing function of the age. For general non-decreasing functions of the age, an optimal sampling policy is designed. The optimal policy is stationary deterministic and has a threshold structure: If the previous transmission was successful, it waits until the age exceeds a threshold and sends out a fresh sample. If the previous transmission failed, it sends out a fresh sample without waiting. The threshold is a root of a fixed-point equation and can be computed with low complexity such as bisection search. Furthermore, we generalize the work by adding a sampling rate constraint. The optimal policy can be randomized but has a similar threshold structure.

 

 

Energy Efficiency in MTC Networks

 

·         We study transmission scheduling with imperfect channel state information (CSI) in a system, where updates are periodically generated and transmitted to the destination over a time-correlated fading            channel. We consider two practical ways to obtain CSI (i.e., delayed sensing and transmission feedback) and investigate the age minimization problem under average energy constraints in the two different cases. We demonstrate that for each case, the optimal policy is a randomization of two threshold-type policies. The threshold-type policy in the case without sensing has a threshold in the belief on good channel state while that in the case with sensing has a threshold in the age. In addition, we propose structure-aware scheduling algorithms for each case.

·         We consider a status update system, where an access point collects samples from multiple sensors and transmits the collected samples to the destination over error-prone channels. Our goal is to minimize the long-term average age with constraints on the long-term average energy and distortion of each update, where the distortion is determined by the number of collected samples. Under assumption that the distortion requirement is a non-decreasing function of the age, we show that the optimal policy is a mixture of two stationary deterministic policies, each of which is optimal for a parameterized average cost problem and is of a threshold-type, i.e., transmission is conducted when the age exceeds a threshold and the distortion requirement is met. We derive the average cost under the threshold-type policy as a piecewise function, and analytically characterize its performance in the large threshold regime. With these, we devise a low-complexity algorithm for the original problem. Moreover, we consider a special case when the distortion requirement is independent of the age, and provide a closed-form solution for the parameterized problem.

 

Age-based Scheduling in MTC Networks

 

·         We consider the age-based scheduling minimization problem, where a source can transmit the status updates over two heterogeneous channels. This work is motivated by recent developments in 5G mmWave technology, in which the mmWave channel is unreliable but fast, and the sub-6GHz channel is slow but reliable. The unreliable channel is modeled as a time correlated Markovian channel at a high data rate. The reliable channel is deterministic with lower data rate. The scheduler decides the channel to be used in each time slot, aiming to minimize the long-term average expected age. It is shown that there exists a multi-dimensional threshold-based policy that is optimal. We provide the optimal thresholds with low complexity (e.g., bisection search) for some portion of the values of the channel parameters. Furthermore, we extend the study by providing for all of the possible values of the channel parameters. To obtain the optimal solution requires overcoming some technical challenges, because super-modularity does not hold over some portion of the state space.

·         We focus on the canonical update-through-queue systems (UTQS), study and quantify the impact of 2-way delay, and aim to design the corresponding optimal age-minimization algorithms. Specifically, the UTQS model considers a typical scenario that the source node would like to send an update packet to the destination node.  Any computation, control, and communication delays experienced by the update packets are lumped together as a basic first-in/first-out queue. The design goal is to adjust the packet injection times so that the updates at the destination are the “freshest” in the long run. The simplicity of such a UTQS model makes it very versatile, with the applications ranging from database update, remote sensing and estimation, cyber-physical system control, cloud-to-edge computing, etc., and is widely recognized as one of the most influential models of age minimization.  Existing results generally fall into two categories: The open-loop setting for which the source is oblivious of the actual packet departure time, versus the closed-loop setting for which the decision is based on instantaneous Acknowledgement (ACK) of when the packet has been served and left the queue. Unfortunately, neither setting accurately reflects modern networked systems, which almost always rely on feedback that experiences some delay. Motivated by this observation, we study the same UTQS model but subject the ACK traffic to an independent queue so that the closed-loop decision can only be made based on delayed feedback. Near-optimal schedulers have been devised that successfully quantifies the impact of 2-way delay on the age minimization problem.

·         We have focused on the study of optimizing the tradeoff between data freshness and update cost in a discrete-time information-update system, aiming to minimize the sum of the staleness cost and the update cost. We first study the tradeoff between the data freshness and the update cost by formulating an optimization problem to minimize the sum of the staleness cost (which is a function of the age-of-information) and the update cost. Then, we provide two useful guidelines for the design of efficient update policies. Following these guidelines and assuming that the aggregated request arrival process is Bernoulli, we prove that there exists a threshold-based policy that is optimal among all online policies and thus focus on the class of threshold-based policies. Furthermore, we derive the closed-form formula for computing the long-term average cost under any threshold-based policy and obtain the optimal threshold. Finally, we perform extensive simulations using both synthetic data and real traces to verify our theoretical results and demonstrate the superior performance of the optimal threshold-based policy compared with several baseline policies.

 

 

Age/quality-based Scheduling for Learning over MTC networks

·         Machine learning over MTC networks have been a driving force for both new machine learning algorithm developments and next generation network designs, with numerous potential future applications in autonomous driving, smart agriculture, and next-generation ML-based communication network solutions. One challenge of distributed network learning systems is the tradeoff between training speed (how to avoid the straggler) and the communication and computation resources (how to distribute the tasks efficiently). In this direction, we focus on the distributed asynchronous gradient descent problems. We observe that if we use the “Age” as the proxy of the quality indicator, we can avoid transmission of low-quality updates and while spending the communication/computation resources only on the high-quality ones. By generalizing the above observation, we have developed a new network communication protocol that directly quantifies the quality of each asynchronous gradient descent update, which reduces the communication overhead, improves the learning convergence rates, stability, accuracy, and reduces the corresponding resource consumption.

 

Resource-scheduling for networks with very large degrees-of-freedom

·         We considered the resource scheduling problem for massive MIMO systems with its optimal solution known to be NP-hard. Massive MIMO systems have a large number of antennas, and hence the systems have a large number of signaling degrees-of-freedom. This research was performed in collaboration with researchers from Nvidia.

·         Inspired by recent achievements in deep reinforcement learning (DRL) to solve problems with large action sets, we developed SMART, a dynamic scheduler for massive MIMO based on the state-of-the-art Soft Actor-Critic (SAC) DRL model and the K-Nearest Neighbors (KNN) algorithm. Through comprehensive simulations using realistic massive MIMO channel models as well as real-world datasets from channel measurement experiments, we demonstrate the effectiveness of our proposed model in various channel conditions. Our results show that our proposed model performs very close to the optimal proportionally fair (Opt-PF) scheduler in terms of spectral efficiency and fairness with more than one order of magnitude lower computational complexity in medium network sizes where Opt-PF is computationally feasible. Our results also show the feasibility and high performance of our proposed scheduler in networks with a large number of users and resource blocks.

 

Broader Impacts

The PIs have continuously focused on disseminating the discovery of this project to international research communities, see the product section below. Wang and Shroff co-supervised a female PhD student in 2022. Shroff recruited a new female PhD student who joined in Autumn 2022. Ji has recruited two new female PhD students who joined in Autumn 2022.

Wang has continuously focused on training undergraduate students and GRAs through research activities, group seminars, and conference presentations. Wang has continuously supervised a Purdue-VIP undergraduate research team since fall 2019, with weekly 90-min meetings dedicated to engaging undergraduate students in graduate-level research on wireless communications.

To further promote student engagement, the VIP teams of spring 2022 to spring 2024 continuously participated in the poster presentation of Purdue Undergraduate Research Conference, an in-person event that have resumed on a semester-basis after the pandemic in 2020. For each semester, there are more than 300 student-led posters participating in-person for this university-wide event. Wang also served as one of the judges for the research talks competition consecutively from spring 2022 to spring 2024, which awarded the top 3 in-person research talks for each of the ten schools of Purdue University. 

Ji has supervised five undergraduate students and two high school students to conduct research. In addition, Ji has actively participated in various educational and outreach activities organized by Center for the Enhancement of Engineering Diversity (CEED) at VT, including TechGirls, Black Engineering Excellence at Virginia Tech (BEE VT), Student Transition to Engineering Program (STEP), C-Tech^2, and Galipatia Slush Rush events.

Ji has served as TPC Co-Chair of AoI Workshop 2022, co-located with IEEE INFOCOM 2022.

Shroff co-chaired the Office of the Under Secretary of Defense Future Directions Workshop on Wireless Communications: “XG and Beyond”. This was a high-profile workshop involving key researchers across Networking, Photonics, and Electronics intended to provide guidance to government agencies about strategic investments in XG networks, which include MTC communications.

PI Shroff and PI Ji, along with several other colleagues, co-organized the 2024 Artificial Intelligence Modeling, Analysis, and Control of Complex Systems (AIMACCS) Workshop, which consists of 13 keynotes, an industry roundtable panel for students, a Women in AI meeting, and a student poster session, where students working on this project presented their research results and progress with AI Edge Institute members and other AIMACCS workshop attendees and received useful feedback.

 

Products

The results in the NSF Public Access Repository will include a comprehensive listing of all journal publications recorded to date that are associated with this award.

Journal papers:

·         G. Yao, A.M. Bedewy, and N. B. Shroff, “Age-optimal low-power status update over time-correlated fading channel,” in IEEE transactions on mobile computing (Early Access), 2022.

·         J. Pan, A. M. Bedewy, Y. Sun, and N. B. Shroff, “Age-optimal Scheduling over Hybrid Channels,” in IEEE Transactions on Mobile Computing (Early Access), 2022. 

·         J. Pan, A. M. Bedewy, Y. Sun, and N. B. Shroff, “Optimal Sampling for Data Freshness: Unreliable Transmissions with Random Two-way Delay,” in IEEE/ACM Transactions on Networking (Early Access), 2022

·         C. -H. Tsai and C. -C. Wang, "Unifying AoI Minimization and Remote Estimation—Optimal Sensor/Controller Coordination With Random Two-Way Delay," in IEEE/ACM Transactions on Networking, vol. 30, no. 1, pp. 229-242, Feb. 2022, doi: 10.1109/TNET.2021.3111495.

·         R. Doost-Mohammady, L. Zhong and A. Sabharwal, “RENEW: A Software-defined Massive MIMO Wireless Experimentation Platform,” ACM GetMobile – Mobile Computing and Computations, 2022.

·         C.-H. Tsai and C.-C. Wang, “Distribution-oblivious Online Algorithms for Age-of-Information Penalty Minimization” IEEE/ACM Transactions on Networking (Early Access), 2023.

·         Z. Liu, B. Li, Z. Zheng, T. Hou, and B. Ji, “Towards Optimal Tradeoff Between Data Freshness and Update Cost in Information-update Systems,” IEEE Internet of Things Journal (IoT-J), 2023. (Early access)

·         S. Kang, A. Eryilmaz, and N.B Shroff, “Remote tracking of distributed dynamic sources over a random access channel with one-bit updates,” IEEE Trans on Network Science and Engineering, Jan. 2023.

·         M. K. C. Shisher, B. Ji, I.-H. Hou, and Y. Sun, “Learning and Communications Co-Design for Remote Inference Systems: Feature Length Selection and Transmission Scheduling,” IEEE Journal on Selected Areas in Information Theory (JSAIT), vol. 4, pp. 524-538, 2023.

·         Q. An, S. Segarra, C. Dick (Nvidia), A. Sabharwal and R. Doost-Mohammady, “A Deep Reinforcement Learning-Based Resource Scheduler for Massive MIMO Networks,” pp. 242-257, IEEE Transactions on Machine Learning in Communications and Networking, September 2023.

·         J. Pan, Y. Sun, and N. B. Shroff, Sampling for Remote Estimation of the Wiener Process over an Unreliable Channel, Performance Evaluation Review, Volume 52, Issue 1, 2023.

·         C.-C. Wang, “Optimal AoI for Systems With Queueing Delay in Both Forward and Backward Directions”, IEEE/ACM Transactions on Networking, accepted for publication. DOI: https://doi.org/10.1109/TNET.2024.3379895

 

 

Conference papers:

·         G. Yao, A.M. Bedewy, and N.B. Shroff, “Age-Optimal Low-Power Status Update over Time-Correlated Fading Channel,” in Proceedings of IEEE International Symposium on Information Theory (ISIT 21’), 2021.

·         G. Yao, A.M. Bedewy, and N.B. Shroff, “Battle between Rate and Error in Minimizing Age of Information,” in Proceedings of the Twenty-second Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing (Mobihoc ’21), 2021, p. 121-130.

·         J. Pan, A. M. Bedewy, Y. Sun, and N. B. Shroff, “Minimizing Age of Information via Scheduling over Heterogeneous Channels,” ACM MobiHoc’21, Shanghai, China, July 2021.

·         J. Pan, A. M. Bedewy, Y. Sun, and N. B. Shroff, “Optimizing Sampling for Data Freshness: Unreliable Transmissions with Random Two-way Delay,” IEEE INFOCOM, May 2022.

·         C.-C. Wang, “How Useful is Delayed Feedback in AoI Minimization - A Study on Systems With Queues in Both Forward and Backward Directions”, in Proceedings of the IEEE Int'l Symp. Information Theory (ISIT), Espoo, Finland, June 26 – July 1, 2022, 6 pages.

·         C.-H. Tsai and C.-C. Wang, “Jointly Minimizing AoI Penalty and Network Cost Among Coexisting Source-Destination Pairs,” in Proceedings of the IEEE Int'l Symp. Information Theory (ISIT), Melbourne, Australia, USA, July 12-20, 2021, virtual conference, 6 pages.

·         Z. Liu, B. Li, Z. Zheng, Y. T. Hou, and B. Ji, “Towards Optimal Tradeoff Between Data Freshness and Update Cost in Information-update Systems,” Proceedings of ICCCN 2022, Virtual Event, July 2022.

·         Z. Liu, K. Zhang, B. Li, Y. Sun, T. Hou, and B. Ji, “Learning-augmented Online Minimization of Age of Information and Transmission Costs,” Proceedings of IEEE INFOCOM 2024, Workshop on Age and Semantics of Information (ASoI), Vancouver, Canada, May 2024.

·         J. Pan, Y. Sun, and N. B. Shroff, “Sampling for Remote Estimation of the Wiener Process over an Unreliable Channel,” ACM SIGMETRICS 2024.

·         W.J. Lee and C.-C. Wang, “AoI-optimal Scheduling for Arbitrary K-channel Update-Through-Queue Systems,” in Proceedings of the IEEE Int'l Symp. Information Theory (ISIT), Athens, Greece, 2024, 6 pages.

 

Book Chapters:

·         B. Li, B. Ji, and A. Eryilmaz, “Age-Efficient Scheduling in Communication Networks,” in book “Age of Information: Foundations and Applications,” Edited by Walid Saad, Harpreet Dhillon, Nikolaos Pappas, Mohamed A. Abd-Elmagid, and Bo Zhou, Cambridge University Press, 2023.

·         A. M. Bedewy, Y. Sun, S. Kompella, and N. B. Shroff, “Sampling and Scheduling for Minimizing Age of Information of Multiple Sources,” Chapter 7, in Age of Information: Foundations and Applications, ISBN 9781108943321, Cambridge University Press, 2023.

 

The results in the NSF Public Access Repository will include a comprehensive listing of all journal publications

 

 

Graduate Research Assistants & Collaborators

Graduate Research Assistants:

·         Zhongdong Liu;

·         Keyuan Zhang;

·         Won Jun Lee;

·         Giles Bischoff;

 

Other Collaborators:

·         I-Hong Hou (Texas A&M University, USA)

·         Y. Thomas Hou (Virginia Tech, USA)

·         Bin Li (University of Rhode Island, USA)

·         Yin Sun (Auburn University, USA)

·         Zizhan Zheng (Tulane University, USA)