More Like this

1. Low-Latency Preamble-Free Transmission of Short Messages via Quickest Change Detection

   Bischoff, G; Wang, C-C (June 2025, IEEE)

   The latency and control overhead of sending the preamble in synchronous communications can be excessive when transmitting short sensing/control messages. To reduce these overheads, this work proposes a preamble-free solution based on the framework of quickest change detection. Specific contributions include a joint decoding/demodulation scheme that is provably asymptotically optimal, and a more practical CuSum-like implementation. Numerical results show that the proposed scheme reduces the latency by 47%–79% when compared to the preamble-based solutions. The scheme is also inherently robust and automatically adapts to any unknown underlying SNRs. 

   more » « less
2. Preamble detection in NB-IoT random access with limited-capacity backhaul

   Ta, Hien; Wang, Zhengdao; Kim, Sang Wu; Nielsen, Jimmy J; Popovski, Petar (January 2019, International Conference on Communications - Proceedings)

   We study multi-base station (BS) preamble detection schemes for the narrow-band Internet of Things (NB-IoT) random access by using stochastic geometry analysis. Specifically, we compare the preamble detection performance of two baseline detection schemes: Quantize-and-Forward (QnF) and Detect-and-Forward (DnF). QnF requires the feedback of quantized received power levels while DnF requires 1-bit feedback of local detection result. Our results show that DnF scheme outperforms QnF scheme when the backhaul capacity is limited or when the minimum distance between user and BSs is less than a threshold. Our results also show that the use of multiple collaborative BSs can lead to a significant improvement of the preamble detection performance, as well as reduction of the total power of the preamble transmission. 

   more » « less
3. OptimML: Joint Control of Inference Latency and Server Power Consumption for ML Performance Optimization

   https://doi.org/10.1145/3661825  

   Chen, Guoyu; Wang, Xiaorui (May 2024, ACM Transactions on Autonomous and Adaptive Systems)

   Power capping is an important technique for high-density servers to safely oversubscribe the power infrastructure in a data center. However, power capping is commonly accomplished by dynamically lowering the server processors’ frequency levels, which can result in degraded application performance. For servers that run important machine learning (ML) applications with Service-Level Objective (SLO) requirements, inference performance such as recognition accuracy must be optimized within a certain latency constraint, which demands high server performance. In order to achieve the best inference accuracy under the desired latency and server power constraints, this paper proposes OptimML, a multi-input-multi-output (MIMO) control framework that jointly controls both inference latency and server power consumption, by flexibly adjusting the machine learning model size (and so its required computing resources) when server frequency needs to be lowered for power capping. Our results on a hardware testbed with widely adopted ML framework (including PyTorch, TensorFlow, and MXNet) show that OptimML achieves higher inference accuracy compared with several well-designed baselines, while respecting both latency and power constraints. Furthermore, an adaptive control scheme with online model switching and estimation is designed to achieve analytic assurance of control accuracy and system stability, even in the face of significant workload/hardware variations. 

   more » « less
4. Embedding security into ferroelectric FET array via in situ memory operation

   https://doi.org/10.1038/s41467-023-43941-5  

   Xu, Yixin; Xiao, Yi; Zhao, Zijian; Müller, Franz; Vardar, Alptekin; Gong, Xiao; George, Sumitha; Kämpfe, Thomas; Narayanan, Vijaykrishnan; Ni, Kai (December 2023, Nature Communications)

   Abstract Non-volatile memories (NVMs) have the potential to reshape next-generation memory systems because of their promising properties of near-zero leakage power consumption, high density and non-volatility. However, NVMs also face critical security threats that exploit the non-volatile property. Compared to volatile memory, the capability of retaining data even after power down makes NVM more vulnerable. Existing solutions to address the security issues of NVMs are mainly based on Advanced Encryption Standard (AES), which incurs significant performance and power overhead. In this paper, we propose a lightweight memory encryption/decryption scheme by exploiting in-situ memory operations with negligible overhead. To validate the feasibility of the encryption/decryption scheme, device-level and array-level experiments are performed using ferroelectric field effect transistor (FeFET) as an example NVM without loss of generality. Besides, a comprehensive evaluation is performed on a 128 × 128 FeFET AND-type memory array in terms of area, latency, power and throughput. Compared with the AES-based scheme, our scheme shows ~22.6×/~14.1× increase in encryption/decryption throughput with negligible power penalty. Furthermore, we evaluate the performance of our scheme over the AES-based scheme when deploying different neural network workloads. Our scheme yields significant latency reduction by 90% on average for encryption and decryption processes. 

   more » « less
5. CSIscan: Learning CSI for Efficient Access Point Discovery in Dense WiFi Networks

   Sankhe, Kunal; Jaisinghani, Dheryta; Chowdhury, Kaushik (October 2020, IEEE International Conference on Network Protocols)

   Network densification through the deployment of WiFi access points (APs) is a promising solution towards achieving high connectivity rates required for emerging applications. A critical first step is to discover an AP before an active association between the client and the AP can be established. Legacy AP discovery procedures initiated by the client result in high latency in the order of a few 100 ms and waste spectrum, especially when clients need to frequently switch between multiple APs. We propose CSIscan that exploits the broadcast nature of WiFi channels by embedding discovery related information within an AP’s ongoing regular transmissions. The AP does this by intelligently distorting the transmitted OFDM frame by inducing perturbations in the preamble, and these injected ‘bits’ of information are detected via changes in the perceived channel state information (CSI). A deep learning framework allocates the optimal level of distortion on a per-subcarrier basis that keeps the resulting packet error rate to less than 1%. Existing clients perceive no changes in their ongoing communication, while potential new clients quickly obtain discovery information at the same time. We experimentally demonstrate that CSIscan reduces the overall WiFi latency from 150 ms to 10 ms and improves spectrum utilization with ∼ 72% reduction in the probe traffic. We show that CSIscan delivers up to 40 discovery information bits in the outgoing WiFi packet in an indoor environment. 

   more » « less