Energy-harvesting (EH) Internet of Things (IoT) devices operate under intermittent energy availability, which disrupts task execution and makes energy-intensive over-the-air (OTA) updates particularly challenging. Conventional OTA update mechanisms rely on reboots and incur significant overhead, rendering them unsuitable for intermittently powered systems. Recent live OTA update techniques reduce reboot overhead but still lack mechanisms to ensure consistency when updates interact with runtime execution. This paper presents AERO, an Adaptive and Efficient Runtime-Aware OTA update mechanism that integrates update tasks into the device’s Directed Acyclic Graph (DAG) and schedules them alongside routine tasks under energy and timing constraints. By identifying update-affected execution regions and dynamically adjusting dependencies, AERO ensures consistent update integration while adapting to intermittent energy availability. Experiments on representative workloads demonstrate improved update reliability and efficiency compared to existing live update approaches.

The widespread deployment of various tiny energy harvesting devices has facilitated the expansion of Internet of Things (IoT) applications, notably in remote and hard-to-reach areas. Once deployed, a critical limitation of these devices is their inability to adapt code to evolving environmental conditions or user requirements. This challenge primarily arises from frequent power interruptions during code updates in energy harvesting devices, unlike their battery-powered counterparts, which can lead to significant errors or system failures. In response, we have designed an innovative framework for facilitating intermittent over-the-air (OTA) code updates in tiny energy harvesting devices. Our approach incorporates Intermittent-aware Update Operations, including insert, modify, delete, and copy, tailored for a variety of update scenarios while accommodating intermittent power and resource constraints. Furthermore, We have designed a Fault-tolerant bootloader that enables the intermittent update capability. This advanced bootloader enables code updates without system reboots and ensures correct task resumption of both routine and update tasks. This not only conserves energy by reducing the need for repetitive reboots but also ensures consistent code updates despite frequent power failures. Additionally, our framework integrates an update-aware checkpointing mechanism to provide reliable backups for both routine tasks and update tasks. This proposed framework presents a general solution for enabling intermittent code updates in tiny energy harvesting devices. Our experimental results demonstrate that the proposed approach outperforms existing approaches under conditions of insufficient harvested energy.

Deep Neural Networks (DNNs) are increasingly applied in various intelligent applications for enhanced accuracy for in-situ decision-making. Considering the cost and longevity, those intelligent applications usually employ energy harvesting (EH) for power supply. Nevertheless, due to inherent intermittency, EH power can frequently disrupt the runtime operation, resulting in subsequent forward progress loss when executing long computations of DNN inference. To address this issue, DNN tiling has been employed where the input data is partitioned into multiple smaller tiles for efficient runtime operation. However, under the energy harvesting scenarios, the size of the tiles can influence runtime energy efficiency significantly under different EH conditions. Therefore, we proposed environmentally adaptive dynamic DNN tiling methods to optimize energy efficiency and runtime reliability. The experimental results on a real testbed show that the proposed technique can outperform the state-of-the-art methods by 19.24% on average.

Energy harvesting (EH) IoT devices that operate intermittently without batteries, coupled with advances in deep neural networks (DNNs), have opened up new opportunities for en-abling sustainable smart applications. Nevertheless, implementing those computation and memory-intensive intelligent algorithms on EH devices is extremely difficult due to the challenges of limited resources and intermittent power supply that causes frequent failures. To address those challenges, this paper proposes a methodology that enables fast deep learning with low-energy accelerators for tiny energy harvesting devices. We first propose RAD, a resource-aware structured DNN training framework, which employs block circulant matrix and structured pruning to achieve high compression for leveraging the advantage of various vector operation accelerators. A DNN implementation method, ACE, is then proposed that employs low-energy accelerators to profit maximum performance with small energy consumption. Finally, we further design FLEX, the system support for inter-mittent computation in energy harvesting situations. Experimental results from three different DNN models demonstrate that RAD, ACE, and FLEX can enable fast and correct inference on energy harvesting devices with up to 4.26X runtime reduction, up to 7. 7X energy reduction with higher accuracy over the state-of-the-art.

IoT devices are increasingly being implemented with neural network models to enable smart applications. Energy harvesting (EH) technology that harvests energy from ambient environment is a promising alternative to batteries for powering those devices due to the low maintenance cost and wide availability of the energy sources. However, the power provided by the energy harvester is low and has an intrinsic drawback of instability since it varies with the ambient environment. This paper proposes EVE, an automated machine learning (autoML) co-exploration framework to search for desired multi-models with shared weights for energy harvesting IoT devices. Those shared models incur significantly reduced memory footprint with different levels of model sparsity, latency, and accuracy to adapt to the environmental changes. An efficient on-device implementation architecture is further developed to efficiently execute each model on device. A run-time model extraction algorithm is proposed that retrieves individual model with negligible overhead when a specific model mode is triggered. Experimental results show that the neural networks models generated by EVE is on average 2.5× times faster than the baseline models without pruning and shared weights.

Deep learning neural networks are of critical importance to enable next-generation IoT devices. However, due to the limited computation power, memory space, and energy, it remains a grand challenge to deploy those algorithms on IoT devices efficiently as they demand high computation, energy, and memory footprint. Numerous pruning methods of deep learning algorithms have been proposed to minimize the latency, energy, and weights. However, few consider the running time memory footprint and the overhead caused by data movement between the volatile memory and non-volatile memory. This paper proposes four novel memory-ware mechanisms for implementing CNN models on self-restrained embedded systems. The proposed techniques maximize the use of high-speed volatile memory and provide three implementation choices to achieve the minimum energy cost, SRAM space usage, and inference latency, as well as a hybrid trade-off choice of the three features. The experimental evaluation compares their energy cost, time latency, and required run-time memory footprint and demonstrates high implementation efficiency.

Being able to learn from complex data with phase information is imperative for many signal processing applications. Today’s real-valued deep neural networks (DNNs) have shown efficiency in latent information analysis but fall short when applied to the complex domain. Deep complex networks (DCN), in contrast, can learn from complex data, but have high computational costs; therefore, they cannot satisfy the instant decision-making requirements of many deployable systems dealing with short observations or short signal bursts. Recent, Binarized Complex Neural Network (BCNN), which integrates DCNs with binarized neural networks (BNN), shows great potential in classifying complex data in real-time. In this paper, we propose a structural pruning based accelerator of BCNN, which is able to provide more than 5000 frames/s inference throughput on edge devices. The high performance comes from both the algorithm and hardware sides. On the algorithm side, we conduct structural pruning to the original BCNN models and obtain 20 × pruning rates with negligible accuracy loss; on the hardware side, we propose a novel 2D convolution operation accelerator for the binary complex neural network. Experimental results show that the proposed design works with over 90% utilization and is able to achieve the inference throughput of 5882 frames/s and 4938 frames/s for complex NIN-Net and ResNet-18 using CIFAR-10 dataset and Alveo U280 Board.