Search

Deep Neural Networks (DNNs) are becoming common in "learning-enabled" time-critical applications such as autonomous driving and robotics. One approach to protect DNN inference from adversarial actions and preserve model privacy/confidentiality is to execute them within trusted enclaves available in modern processors. However, running DNN inference inside limited-capacity enclaves while ensuring timing guarantees is challenging due to (a) large size of DNN workloads and (b) extra switching between "normal" and "trusted" execution modes. This paper introduces new time-aware scheduling schemes - DeepTrust^RT - to securely execute deep neural inferences for learning-enabled real-time systems. We first propose a variant of EDF (called DeepTrust^RT-LW) that slices each DNN layer and runs them sequentially in the enclave. However, due to extra context switch overheads of individual layer slices, we further introduce a novel layer fusion technique (named DeepTrust^RT-FUSION). Our proposed scheme provides hard real-time guarantees by fusing multiple layers of DNN workload from multiple tasks; thus allowing them to fit and run concurrently within the enclaves while maintaining real-time guarantees. We implemented and tested DeepTrust^RT ideas on the Raspberry Pi platform running OP-TEE+DarkNet-TZ DNN APIs and three DNN workloads (AlexNet-squeezed, Tiny Darknet, YOLOv3-tiny). Compared to the layer-wise partitioning approach (DeepTrust^RT-LW), DeepTrust^RT-FUSION can schedule up to 3x more tasksets and reduce context switches by up to 11.12x. We further demonstrate the efficacy of DeepTrust^RT using a flight controller (ArduPilot) case study and find that DeepTrust^RT-FUSION retains real-time guarantees where DeepTrust^RT-LW becomes unschedulable.

A Hybrid LiFi and WiFi network (HLWNet) integrates the rapid data transmission capabilities of Light Fidelity (LiFi) with the extensive connectivity provided by Wireless Fidelity (WiFi), resulting in significant benefits for wireless data transmissions in the designated area. However, the challenge of decision-making during the handover process in HLWNet is made more complex due to the specific characteristics of electromagnetic signals’ line-of-sight transmission, resulting in a greater level of intricacy compared to previous heterogeneous networks. This research work addresses the problem of handover decisions in the Hybrid LiFi and WiFi networks and treats it as a binary classification problem. Consequently, it proposes a handover method based on a deep neural network (DNN). The comprehensive handover scheme incorporates two sets of neural networks (ANN and DNN) that utilize input factors such as channel quality and the mobility of users to enable informed decisions during handovers. Following training with labeled datasets, the neural-network-based handover approach achieves an accuracy rate exceeding 95%. A comparative analysis of the proposed scheme against the benchmark reveals that the proposed method considerably increases user throughput by approximately 18.58% to 38.5% while reducing the handover rate by approximately 55.21% to 67.15% compared to the benchmark artificial neural network (ANN); moreover, the proposed method demonstrates robustness in the face of variations in user mobility and channel conditions.

A Hybrid LiFi and WiFi network (HLWNet) integrates the rapid data transmission capabilities of Light Fidelity (LiFi) with the extensive connectivity provided by Wireless Fidelity (WiFi), resulting in significant benefits for wireless data transmissions in the designated area. However, the challenge of decision-making during the handover process in HLWNet is made more complex due to the specific characteristics of electromagnetic signals’ line-of-sight transmission, resulting in a greater level of intricacy compared to previous heterogeneous networks. This research work addresses the problem of handover decisions in the Hybrid LiFi and WiFi networks and treats it as a binary classification problem. Consequently, it proposes a handover method based on a deep neural network (DNN). The comprehensive handover scheme incorporates two sets of neural networks (ANN and DNN) that utilize input factors such as channel quality and the mobility of users to enable informed decisions during handovers. Following training with labeled datasets, the neural-network-based handover approach achieves an accuracy rate exceeding 95%. A comparative analysis of the proposed scheme against the benchmark reveals that the proposed method considerably increases user throughput by approximately 18.58% to 38.5% while reducing the handover rate by approximately 55.21% to 67.15% compared to the benchmark artificial neural network (ANN); moreover, the proposed method demonstrates robustness in the face of variations in user mobility and channel conditions.

A Hybrid LiFi and WiFi network (HLWNet) integrates the rapid data transmission capabilities of Light Fidelity (LiFi) with the extensive connectivity provided by Wireless Fidelity (WiFi), resulting in significant benefits for wireless data transmissions in the designated area. However, the challenge of decision-making during the handover process in HLWNet is made more complex due to the specific characteristics of electromagnetic signals’ line-of-sight transmission, resulting in a greater level of intricacy compared to previous heterogeneous networks. This research work addresses the problem of handover decisions in the Hybrid LiFi and WiFi networks and treats it as a binary classification problem. Consequently, it proposes a handover method based on a deep neural network (DNN). The comprehensive handover scheme incorporates two sets of neural networks (ANN and DNN) that utilize input factors such as channel quality and the mobility of users to enable informed decisions during handovers. Following training with labeled datasets, the neural-network-based handover approach achieves an accuracy rate exceeding 95%. A comparative analysis of the proposed scheme against the benchmark reveals that the proposed method considerably increases user throughput by approximately 18.58% to 38.5% while reducing the handover rate by approximately 55.21% to 67.15% compared to the benchmark artificial neural network (ANN); moreover, the proposed method demonstrates robustness in the face of variations in user mobility and channel conditions.

Large Language Models (LLMs) have demonstrated remarkable language understanding and generation capabilities. However, training, deploying, and accessing these models pose notable challenges, including resource-intensive demands, extended training durations, and scalability issues. To address these issues, we introduce a concept of hierarchical, distributed LLM architecture that aims at enhancing the accessibility and deployability of LLMs across heterogeneous computing platforms, including general-purpose computers (e.g., laptops) and IoT-style devices (e.g., embedded systems). By introducing a "layered" approach, the proposed architecture enables on-demand accessibility to LLMs as a customizable service. This approach also ensures optimal trade-offs between the available computational resources and the user's application needs. We envision that the concept of hierarchical LLM will empower extensive, crowd-sourced user bases to harness the capabilities of LLMs, thereby fostering advancements in AI technology in general.

Interfacial electron-transfer (ET) reactions underpin the interconversion of electrical and chemical energy. It is known that the electronic state of electrodes strongly influences ET rates because of differences in the electronic density of states (DOS) across metals, semimetals, and semiconductors. Here, by controlling interlayer twists in well-defined trilayer graphene moirés, we show that ET rates are strikingly dependent on electronic localization in each atomic layer and not the overall DOS. The large degree of tunability inherent to moiré electrodes leads to local ET kinetics that range over 3 orders of magnitude across different constructions of only three atomic layers, even exceeding rates at bulk metals. Our results demonstrate that beyond the ensemble DOS, electronic localization is critical in facilitating interfacial ET, with implications for understanding the origin of high interfacial reactivity typically exhibited by defects at electrode-electrolyte interfaces.

This paper presents a performance study of transformer language models under different hardware configurations and accuracy requirements and derives empirical observations about these resource/accuracy trade-offs. In particular, we study how the most commonly used BERT-based language models (viz., BERT, RoBERTa, DistilBERT, and TinyBERT) perform on embedded systems. We tested them on four off-the-shelf embedded platforms (Raspberry Pi, Jetson, UP2, and UDOO) with 2 GB and 4 GB memory (i.e., a total of eight hardware configurations) and four datasets (i.e., HuRIC, GoEmotion, CoNLL, WNUT17) running various NLP tasks. Our study finds that executing complex NLP tasks (such as "sentiment" classification) on embedded systems is feasible even without any GPUs (e.g., Raspberry Pi with 2 GB of RAM). Our findings can help designers understand the deployability and performance of transformer language models, especially those based on BERT architectures.

Tailoring electron transfer dynamics across solid-liquid interfaces is fundamental to the interconversion of electrical and chemical energy. Stacking atomically thin layers with a small azimuthal misorientation to produce moiré superlattices enables the controlled engineering of electronic band structures and the formation of extremely flat electronic bands. Here, we report a strong twist-angle dependence of heterogeneous charge transfer kinetics at twisted bilayer graphene electrodes with the greatest enhancement observed near the 'magic angle' (~1.1°). This effect is driven by the angle-dependent tuning of moiré-derived flat bands that modulate electron transfer processes with the solution-phase redox couple. Combined experimental and computational analysis reveals that the variation in electrochemical activity with moiré angle is controlled by a structural relaxation of the moiré superlattice at twist angles of <2°, and 'topological defect' AA stacking regions, where flat bands are localized, produce a large anomalous local electrochemical enhancement that cannot be accounted for by the elevated local density of states alone.

Localization of sensor nodes in the Internet of Underwater Things (IoUT) is of considerable significance due to its various applications, such as navigation, data tagging, and detection of underwater objects. Therefore, in this paper, we propose a hybrid Bayesian multidimensional scaling (BMDS) based localization technique that can work on a fully hybrid IoUT network where the nodes can communicate using either optical, magnetic induction, and acoustic technologies. These technologies are already used for communication in the underwater environment; however, lacking localization solutions. Optical and magnetic induction communication achieves higher data rates for short communication. On the contrary, acoustic waves provide a low data rate for long-range underwater communication. The proposed method collectively uses optical, magnetic induction, and acoustic communication-based ranging to estimate the underwater sensor nodes' final locations. Moreover, we also analyze the proposed scheme by deriving the hybrid Cramer Rao lower bound (HCRLB). Simulation results provide a complete comparative analysis of the proposed method with the literature., Comment: Accepted for Pulication in IEEE/CAA Journal of Automatica Sinica