This study describes the Computing Platforms (CPs) and the hardware reliability issues of Unmanned Aerial Vehicles (UAVs), or drones, which recently attracted significant attention in mission and safety-critical applications demanding a failure-free operation. While the rapid development of the UAV technologies was recently reviewed by survey reports focusing on the architecture, cost, energy efficiency, communication, and civil application aspects, the computing platforms’ reliability perspective was overlooked. Moreover, due to the rising complexity and diversity of today’s UAV CPs, their reliability is becoming a prominent issue demanding up-to-date solutions tailored to the UAV specifics. The objective of this work is to address this gap, focusing on the hardware reliability aspect. This research studies the UAV CPs deployed for representative applications, specific fault and failure modes, and existing approaches for reliability assessment and enhancement in CPs for failure-free UAV operation. This study indicates how faults and failures occur in the various system layers of UAVs and analyzes open challenges. We advocate a concept of a cross-layer reliability model tailored to UAVs’ onboard intelligence and identify directions for future research in this area.
14 September 2020 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
In this study, a novel method for the detection of recycled field-programmable gate arrays (FPGAs) is proposed. This method is based on with-in die (WID) process variation model over an exhaustive path characterization [referred to as exhaustive-fingerprint (X-FP)]. In the proposed method, X-FP is capable of fully characterizing frequencies on all paths in look-up tables (LUTs) using advanced ring oscillator (RO) design to exhaustively capture deterioration by aging. Although machine learning (ML)-based classification is often used for recycled FPGA detection, X-FP yields a large amount of measurement data, which cannot be appropriately handled by typical ML algorithms, if they are used as a feature vector. The proposed method utilizes the model parameters extracted by WID variation modeling as the feature vector in the ML algorithm. These model parameters simply and accurately represent the process variation for each FPGA. In this way, the ML-based fresh/recycled classification works very well with the simple feature vector. Experiments using 50 commercially available FPGAs reveal that X-FP can capture the degradation effects, which cannot be detected by conventional methods. Moreover, the WID modeling achieves 99.6% feature size reduction per one FPGA. It also demonstrates that the model parameters exhibit good distance properties between fresh and aged FPGAs. Additionally, the ML-based classification which uses a one-class support vector machine can successfully detect 2 aged FPGAs (48 h accelerated aging) without any misclassification and another 4 aged FPGAs (24 h accelerated aging) with a very few misclassifications as fresh FPGAs.
Analyzing aging-induced delay degradations of ring oscillators (ROs) is an effective way to detect recycled fieldprogrammable gate arrays (FPGAs). On the other hand, it requires a large number of measurements of ROs for all FPGAs before shipping, and thus leads to measurement cost inflation. In this research, we propose a low-cost recycled FPGA detection method using a virtual probe (VP) technique based on compressed sensing. The VP technique enables us to accurately predict the spatial process variation on a die from a very small number of sample measurements. Using the estimated process variation as a supervisor, machine-learning algorithm classifies target FPGAs into recycled or fresh. Through experiments using circuit simulation, our method achieves more than 96% detection accuracy using one-class support vector machine where only 20% samples of the frequency are used at the best case. Silicon measurement results on Xilinx Artix-7 FPGAs also demonstrate the efficiencies of the proposed method.
The rapid advancement of autonomous and heterogeneous unmanned aerial vehicle (UAV) swarms necessitates efficient computation offloading (CO) strategies to optimize their performance in industries, e.g., disaster management, surveillance, and environmental monitoring. UAVs face constraints such as limited energy, latency requirements, and failure risks, making robust CO approaches essential. Current CO methods often fall short due to high energy consumption, increased latency, and reliability issues in challenging conditions. This work introduces a novel collaborative CO strategy to address these deficiencies. Our approach utilizes a Bayesian network for failure mode effect analysis, considering communication bit error probabilities among multiantenna UAVs. We further employ rating-based federated deep learning to optimize decision-making, determining the best CO destination for each UAV based on factors like positions and resource capacities. Our strategy significantly outperforms existing benchmarks and state-of-the-art methods. It decreases the average probability of critical task failure by 43% and reduces energy consumption by 15% on average ensuring UAV swarms can meet strict constraints in harsh environments. These improvements demonstrate the utility of our approach in enhancing the operational reliability and efficiency of UAV swarms across diverse applications.
We propose a novel recycled field programmable gate array (FPGA) detection method based on with-in die (WID) variation modeling. Model parameters extracted by the WID modeling simply but accurately represent the process variation of each FPGA, and thus, by using the model parameters as feature vector, machine learning effectively classifies target FPGAs into fresh or aged. Through experiment on commercial circuit simulator, the proposed method achieves 99.6% feature vector size reduction per FPGA with showing 96.0% detection accuracy; while our silicon result using commercial FPGA demonstrates the model parameters have a good distance properties between fresh and aged.