[{"id":"benchmarking-ultra-low-power-munpus","title":"Benchmarking Ultra-Low-Power μNPUs","short_name":null,"former_titles":[],"authors":["Josh Millar","Yushan Huang","Sarab Sethi","Hamed Haddadi","Anil Madhavapeddy"],"year":2025,"publication_date":"2025-11-04","venue":{"full":"Proceedings of the 31st Annual International Conference on Mobile Computing and Networking","short":"ACM MobiCom 2025"},"type":"conference","publisher":"Association for Computing Machinery","volume":null,"issue":null,"pages":"1060-1074","doi":"10.1145/3680207.3765264","arxiv":"2503.22567","publisher_url":"https://dl.acm.org/doi/10.1145/3680207.3765264","pdf_url":"https://arxiv.org/pdf/2503.22567","code_url":"https://github.com/j0shmillar/uNPU-Bench","abstract":"Efficient on-device neural network (NN) inference offers predictable latency, improved privacy and reliability, and lower operating costs for vendors than cloud-based inference. This has sparked recent development of microcontroller-scale NN accelerators, also known as neural processing units (μNPUs), designed specifically for ultra-low-power applications. We present the first comparative evaluation of a number of commercially-available μNPUs, including the first independent benchmarks for multiple platforms. To ensure fairness, we develop and open-source a model compilation pipeline supporting consistent benchmarking of quantized models across diverse microcontroller hardware. Our resulting analysis uncovers both expected performance trends as well as surprising disparities between hardware specifications and actual performance, including certain μNPUs exhibiting unexpected scaling behaviors with model complexity. This work provides a foundation for ongoing evaluation of μNPU platforms, alongside offering practical insights for both hardware and software developers in this rapidly evolving space.","problem":"Commercial microcontroller-scale NPUs are difficult to compare because vendor measurements use different models, quantization schemes, toolchains, and definitions of inference cost. This paper asks how these platforms compare under consistent workloads when end-to-end latency, power, memory transfers, initialization, and CPU post-processing are measured together.","contributions":["A side-by-side evaluation of commercially available μNPU platforms under consistent workloads and measurement conditions.","Independent, fine-grained benchmarks for platforms that previously lacked third-party evaluation.","An open-source compilation toolchain for deploying equivalent quantized neural networks across diverse μNPU hardware.","Practical platform-selection and model-optimization recommendations derived from stage-by-stage measurements."],"key_results":["Dedicated neural accelerators achieved up to two orders of magnitude higher energy efficiency than general-purpose MCUs in the evaluated workloads.","The HX-WE2 averaged about 1.93× lower end-to-end latency than the MAX78000, with about 3.13× higher power consumption.","The MAX78000 spent up to 90% of execution time on memory I/O, showing why advertised GOPS alone did not predict end-to-end performance."],"topics":["TinyML","Neural Processing Units","Microcontrollers","Embedded AI","Hardware Benchmarking","Energy-Efficient Inference","Model Compilation"],"bibtex":"@inproceedings{millar2025benchmarking,\n  author    = {Josh Millar and Yushan Huang and Sarab Sethi and Hamed Haddadi and Anil Madhavapeddy},\n  title     = {Benchmarking Ultra-Low-Power {$\\mu$NPUs}},\n  booktitle = {Proceedings of the 31st Annual International Conference on Mobile Computing and Networking},\n  year      = {2025},\n  pages     = {1060--1074},\n  doi       = {10.1145/3680207.3765264}\n}"},{"id":"energy-characterization-tiny-ai-accelerator-mcus","title":"Energy Characterization of Tiny AI Accelerator-Equipped Microcontrollers","short_name":null,"former_titles":[],"authors":["Yushan Huang","Taesik Gong","SiYoung Jang","Fahim Kawsar","Chulhong Min"],"year":2024,"publication_date":"2024-11-04","venue":{"full":"Proceedings of the 2nd International Workshop on Human-Centered Sensing, Networking, and Multi-Device Systems","short":"HumanSys 2024"},"type":"workshop","publisher":"Association for Computing Machinery","volume":null,"issue":null,"pages":"1-6","doi":"10.1145/3698388.3699628","arxiv":null,"publisher_url":"https://dl.acm.org/doi/10.1145/3698388.3699628","pdf_url":"https://chulhong.github.io/paper/3698388.3699628.pdf","code_url":null,"award":"Best Paper Award","abstract":"Tiny AI accelerators are seamlessly integrated into wearable devices due to their small form factor, enabling human sensing applications to run solely on wearables. However, despite this potential, the energy characterization of these tiny AI accelerators has been hardly studied, which is a key enabler for realizing such applications in our daily lives. In this paper, we present a comprehensive analysis of the energy characterization of ultra-low power microcontrollers using MAX78000 manufactured by Analog Device. We detailed the hardware components and their supported power configurations. We then conducted extensive benchmarks at micro and macro levels. For micro-level benchmarks, we evaluated the power/energy consumption under individual system configuration involved in each operation—sensing, AI inference, computation, memory I/O, and idle. For macro-level benchmarks, we analyzed the impact of system-wide configurations on overall energy consumption of end-to-end application pipelines. Our findings offer valuable insights into energy optimization for wearable systems with on-device and human-centered sensing technologies.","problem":"Tiny AI accelerators make fully on-wearable sensing and inference possible, but their energy behavior across hardware components, power modes, and end-to-end application stages was not well characterized. This makes it difficult to choose configurations that preserve battery life in continuous human-centered sensing.","contributions":["A component-level account of the MAX78000 hardware and its supported power configurations.","Microbenchmarks of sensing, AI inference, general computation, memory I/O, and idle operation.","Macrobenchmarks showing how system-wide configurations affect complete sensing-and-inference pipelines.","Energy-optimization guidance for wearable, on-device sensing systems."],"key_results":["The measurements show that application energy depends on both accelerator execution and surrounding operations such as sensing, memory I/O, computation, and idle configuration."],"topics":["Tiny AI Accelerators","Microcontrollers","Wearable Computing","Human-Centered Sensing","Energy Characterization","On-Device Inference","MAX78000"],"bibtex":"@inproceedings{huang2024energy,\n  author    = {Yushan Huang and Taesik Gong and SiYoung Jang and Fahim Kawsar and Chulhong Min},\n  title     = {Energy Characterization of Tiny {AI} Accelerator-Equipped Microcontrollers},\n  booktitle = {Proceedings of the 2nd International Workshop on Human-Centered Sensing, Networking, and Multi-Device Systems},\n  year      = {2024},\n  pages     = {1--6},\n  doi       = {10.1145/3698388.3699628}\n}"},{"id":"microt","title":"Low-Energy On-Device Personalization for MCUs","short_name":"MicroT","former_titles":["Low-Energy and Adaptive Models for MCUs"],"authors":["Yushan Huang","Ranya Aloufi","Xavier Cadet","Yuchen Zhao","Payam Barnaghi","Hamed Haddadi"],"year":2024,"publication_date":"2024-12-04","venue":{"full":"2024 IEEE/ACM Symposium on Edge Computing","short":"SEC 2024"},"type":"conference","publisher":"IEEE","volume":null,"issue":null,"pages":"45-58","doi":"10.1109/SEC62691.2024.00012","arxiv":"2403.08040","publisher_url":"https://ieeexplore.ieee.org/document/10818186","pdf_url":"https://arxiv.org/pdf/2403.08040","code_url":"https://github.com/yushan-huang/MicroT","abstract":"Microcontroller Units (MCUs) are ideal platforms for edge applications due to their low cost and energy consumption, and are widely used in various applications, including personalized machine learning tasks, where customized models can enhance the task adaptation. However, existing approaches for local on-device personalization mostly support simple ML architectures or require complex local pre-training/training, leading to high energy consumption and negating the low-energy advantage of MCUs. In this paper, we introduce MicroT, an efficient and low-energy MCU personalization approach. MicroT includes a robust, general, but tiny feature extractor, developed through self-supervised knowledge distillation, which trains a task-specific head to enable independent on-device personalization with minimal energy and computational requirements. MicroT implements an MCU-optimized early-exit inference mechanism called stage-decision to further reduce energy costs. This mechanism allows for user-configurable exit criteria (stage-decision ratio) to adaptively balance energy cost with model performance. We evaluated MicroT using two models, three datasets, and two MCU boards. MicroT outperforms traditional transfer learning (TTL) and two SOTA approaches by 2.12–11.60% across two models and three datasets. Targeting widely used energy-aware edge devices, MicroT's on-device training requires no additional complex operations, halving the energy cost compared to SOTA approaches by up to 2.28× while keeping SRAM usage below 1 MB. During local inference, MicroT reduces energy cost by 14.17% compared to TTL across two boards and two datasets, highlighting its suitability for long-term use on energy-aware resource-constrained MCUs.","problem":"Personalization helps edge models adapt to a user's task or data, but existing on-device methods either support only simple models or impose training costs that erase the energy advantage of MCU deployment. MicroT targets independent local personalization under tight SRAM, compute, and energy budgets.","contributions":["A tiny, general feature extractor trained with self-supervised knowledge distillation so an MCU can personalize only a task-specific head.","A stage-decision early-exit mechanism with a configurable ratio for trading model performance against inference energy.","An MCU-oriented training path that avoids additional complex local operations and keeps SRAM use below 1 MB.","Evaluation across two models, three datasets, and two MCU boards."],"key_results":["MicroT improved performance by 2.12–11.60% over traditional transfer learning and two compared approaches across the evaluated models and datasets.","On-device training used up to 2.28× less energy than the compared approaches while keeping SRAM usage below 1 MB.","Local inference reduced energy cost by 14.17% compared with traditional transfer learning across two boards and two datasets."],"topics":["TinyML","On-Device Learning","Model Personalization","Microcontrollers","Self-Supervised Knowledge Distillation","Early-Exit Inference","Energy-Efficient Machine Learning","Edge AI"],"bibtex":"@inproceedings{huang2024microt,\n  author    = {Yushan Huang and Ranya Aloufi and Xavier Cadet and Yuchen Zhao and Payam Barnaghi and Hamed Haddadi},\n  title     = {Low-Energy On-Device Personalization for {MCUs}},\n  booktitle = {2024 IEEE/ACM Symposium on Edge Computing (SEC)},\n  year      = {2024},\n  pages     = {45--58},\n  doi       = {10.1109/SEC62691.2024.00012}\n}"},{"id":"tbft","title":"Towards Low-Energy Adaptive Personalization for Resource-Constrained Devices","short_name":"TBFT","former_titles":[],"authors":["Yushan Huang","Josh Millar","Yuxuan Long","Yuchen Zhao","Hamed Haddadi"],"year":2024,"publication_date":"2024-04-22","venue":{"full":"Proceedings of the 4th Workshop on Machine Learning and Systems","short":"EuroMLSys 2024"},"type":"workshop","publisher":"Association for Computing Machinery","volume":null,"issue":null,"pages":"73-80","doi":"10.1145/3642970.3655826","arxiv":"2403.15905","publisher_url":"https://dl.acm.org/doi/10.1145/3642970.3655826","pdf_url":"https://arxiv.org/pdf/2403.15905","code_url":"https://github.com/yushan-huang/AdaptivePersonalization","abstract":"The personalization of machine learning (ML) models to address data drift is a significant challenge in the context of Internet of Things (IoT) applications. Presently, most approaches focus on fine-tuning either the full base model or its last few layers to adapt to new data, while often neglecting energy costs. However, various types of data drift exist, and fine-tuning the full base model or the last few layers may not result in optimal performance in certain scenarios. We propose Target Block Fine-Tuning (TBFT), a low-energy adaptive personalization framework designed for resource-constrained devices. We categorize data drift and personalization into three types: input-level, feature-level, and output-level. For each type, we fine-tune different blocks of the model to achieve optimal performance with reduced energy costs. Specifically, input-, feature-, and output-level correspond to fine-tuning the front, middle, and rear blocks of the model. We evaluate TBFT on a ResNet model, three datasets, three different training sizes, and a Raspberry Pi. Compared with BlockAvg, where each block is fine-tuned individually and their performance improvements are averaged, TBFT exhibits an improvement in model accuracy by an average of 15.30% whilst saving 41.57% energy consumption on average compared with full fine-tuning.","problem":"Distribution shifts can occur at the input, feature, or output level, yet common personalization methods always fine-tune the full network or its final layers. That fixed choice can waste energy and adapt the wrong part of a model on a resource-constrained edge device.","contributions":["A three-part taxonomy linking input-, feature-, and output-level drift to front, middle, and rear model blocks.","Target Block Fine-Tuning, which chooses the model region to update according to the detected personalization scenario.","An evaluation spanning a ResNet, three datasets, three training-set sizes, and Raspberry Pi energy measurements."],"key_results":["TBFT improved accuracy by 15.30% on average relative to the BlockAvg comparison.","TBFT saved 41.57% energy on average relative to full-model fine-tuning."],"topics":["On-Device Personalization","Data Drift","Targeted Fine-Tuning","Resource-Constrained Devices","Edge AI","Energy-Efficient Learning","TinyML"],"bibtex":"@inproceedings{huang2024tbft,\n  author    = {Yushan Huang and Josh Millar and Yuxuan Long and Yuchen Zhao and Hamed Haddadi},\n  title     = {Towards Low-Energy Adaptive Personalization for Resource-Constrained Devices},\n  booktitle = {Proceedings of the 4th Workshop on Machine Learning and Systems},\n  year      = {2024},\n  pages     = {73--80},\n  doi       = {10.1145/3642970.3655826}\n}"},{"id":"entropy-features-neurological-conditions","title":"Analyzing entropy features in time-series data for pattern recognition in neurological conditions","short_name":"Entropy Pipeline","former_titles":[],"authors":["Yushan Huang","Yuchen Zhao","Alexander Capstick","Francesca Palermo","Hamed Haddadi","Payam Barnaghi"],"year":2024,"publication_date":"2024-04","venue":{"full":"Artificial Intelligence in Medicine","short":"Artificial Intelligence in Medicine"},"type":"journal","publisher":"Elsevier","volume":"150","issue":null,"pages":"102821","doi":"10.1016/j.artmed.2024.102821","arxiv":null,"publisher_url":"https://www.sciencedirect.com/science/article/pii/S0933365724000630","pdf_url":null,"code_url":"https://github.com/yushan-huang/EntropyPipeline","abstract":"In the field of medical diagnosis and patient monitoring, effective pattern recognition in neurological time-series data is essential. Traditional methods predominantly based on statistical or probabilistic learning and inference often struggle with multivariate, multi-source, state-varying, and noisy data while also posing privacy risks due to excessive information collection and modeling. Furthermore, these methods often overlook critical statistical information, such as the distribution of data points and inherent uncertainties. To address these challenges, we introduce an information theory-based pipeline that leverages specialized features to identify patterns in neurological time-series data while minimizing privacy risks. We incorporate various entropy methods based on the characteristics of different scenarios and entropy. For stochastic state transition applications, we incorporate Shannon's entropy, entropy rates, entropy production, and the von Neumann entropy of Markov chains. When state modeling is impractical, we select and employ approximate entropy, increment entropy, dispersion entropy, phase entropy, and slope entropy. The pipeline's effectiveness and scalability are demonstrated through pattern analysis in a dementia care dataset and also an epileptic and a myocardial infarction dataset. The results indicate that our information theory-based pipeline can achieve average performance improvements across various models on the recall rate, F1 score, and accuracy by up to 13.08 percentage points, while enhancing inference efficiency by reducing the number of model parameters by an average of 3.10 times. Thus, our approach opens a promising avenue for improved, efficient, and critical statistical information-considered pattern recognition in medical time-series data.","problem":"Neurological and physiological time series are noisy, multivariate, and state-varying. Raw-data learning can miss distributional uncertainty, require unnecessarily detailed personal data, and produce large inference models; the paper investigates whether entropy-derived summaries can provide compact, informative features instead.","contributions":["An information-theoretic pipeline that selects entropy features according to whether a time series can be represented by stochastic state transitions.","Coverage of Shannon entropy, entropy rate, entropy production, von Neumann entropy, and five non-state-based entropy measures.","Evaluation on dementia-care, epilepsy, and myocardial-infarction datasets to test effectiveness across different medical time series.","Analysis of predictive performance, parameter count, and the privacy benefit of working with summarized features rather than excessive raw information."],"key_results":["Across the evaluated models, entropy features improved recall, F1 score, and accuracy by up to 13.08 percentage points on average.","The resulting inference models used 3.10× fewer parameters on average."],"topics":["Medical Time Series","Entropy Features","Information Theory","Neurological Conditions","Dementia Monitoring","Epilepsy","Pattern Recognition","Privacy-Aware Health AI"],"bibtex":"@article{huang2024entropy,\n  author  = {Yushan Huang and Yuchen Zhao and Alexander Capstick and Francesca Palermo and Hamed Haddadi and Payam Barnaghi},\n  title   = {Analyzing entropy features in time-series data for pattern recognition in neurological conditions},\n  journal = {Artificial Intelligence in Medicine},\n  volume  = {150},\n  pages   = {102821},\n  year    = {2024},\n  doi     = {10.1016/j.artmed.2024.102821}\n}"},{"id":"battery-free-ml-mcus","title":"Poster: Towards Battery-Free Machine Learning Inference and Model Personalization on MCUs","short_name":null,"former_titles":["Poster: Towards Machine Learning and Inference for Resource-constrained MCUs"],"authors":["Yushan Huang","Hamed Haddadi"],"year":2023,"publication_date":"2023-06-18","venue":{"full":"Proceedings of the 21st Annual International Conference on Mobile Systems, Applications and Services","short":"ACM MobiSys 2023"},"type":"poster","publisher":"Association for Computing Machinery","volume":null,"issue":null,"pages":"571-572","doi":"10.1145/3581791.3597371","arxiv":null,"publisher_url":"https://dl.acm.org/doi/10.1145/3581791.3597371","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3581791.3597371","code_url":null,"abstract":"Machine learning (ML) is moving towards edge devices. However, ML models with high computational demands and energy consumption pose challenges for ML inference in resource-constrained environments, such as the deep sea. To address these challenges, we propose a battery-free ML inference and model personalization pipeline for microcontroller units (MCUs). As an example, we performed fish image recognition in the ocean. We evaluated and compared the accuracy, runtime, power, and energy consumption of the model before and after optimization. The results demonstrate that our pipeline can achieve 97.78% accuracy with 483.82 KB Flash, 70.32 KB RAM, 118 ms runtime, 4.83 mW power, and 0.57 mJ energy consumption on MCUs, reducing by 64.17%, 12.31%, 52.42%, 63.74%, and 82.67%, compared to the baseline. The results indicate the feasibility of battery-free ML inference on MCUs.","problem":"Remote environments such as the deep sea can make battery replacement or network-dependent inference impractical. This poster studies whether image recognition and personalization can fit within the memory, latency, power, and harvested-energy constraints of an MCU-based battery-free system.","contributions":["A battery-free MCU pipeline combining machine-learning inference with model personalization.","A fish-image recognition case study for a remote ocean sensing scenario.","Joint measurement of accuracy, flash and RAM use, runtime, power, and per-inference energy before and after optimization."],"key_results":["The optimized MCU pipeline achieved 97.78% accuracy using 483.82 KB flash and 70.32 KB RAM.","It ran in 118 ms at 4.83 mW and consumed 0.57 mJ per inference.","Compared with the baseline, the reported reductions were 64.17% in flash, 12.31% in RAM, 52.42% in runtime, 63.74% in power, and 82.67% in energy."],"topics":["Battery-Free Computing","TinyML","Microcontrollers","On-Device Inference","Model Personalization","Energy Harvesting","Embedded Computer Vision"],"bibtex":"@inproceedings{huang2023batteryfree,\n  author    = {Yushan Huang and Hamed Haddadi},\n  title     = {Poster: Towards Battery-Free Machine Learning Inference and Model Personalization on {MCUs}},\n  booktitle = {Proceedings of the 21st Annual International Conference on Mobile Systems, Applications and Services},\n  year      = {2023},\n  pages     = {571--572},\n  doi       = {10.1145/3581791.3597371}\n}"},{"id":"entropy-activity-patterns","title":"Using Entropy Measures for Monitoring the Evolution of Activity Patterns","short_name":null,"former_titles":[],"authors":["Yushan Huang","Yuchen Zhao","Hamed Haddadi","Payam Barnaghi"],"year":2022,"publication_date":"2022-10-26","venue":{"full":"2022 IEEE 8th World Forum on Internet of Things","short":"IEEE WF-IoT 2022"},"type":"conference","publisher":"IEEE","volume":null,"issue":null,"pages":"1-6","doi":"10.1109/WF-IoT54382.2022.10152050","arxiv":"2210.01736","publisher_url":"https://ieeexplore.ieee.org/document/10152050","pdf_url":"https://arxiv.org/pdf/2210.01736","code_url":null,"abstract":"In this work, we apply information theory inspired methods to quantify changes in daily activity patterns. We use in-home movement monitoring data and show how they can help indicate the occurrence of healthcare-related events. Three different types of entropy measures namely Shannon's entropy, entropy rates for Markov chains, and entropy production rate have been utilised. The measures are evaluated on a large-scale in-home monitoring dataset that has been collected within our dementia care clinical study. The study uses Internet of Things (IoT) enabled solutions for continuous monitoring of in-home activity, sleep, and physiology to develop care and early intervention solutions to support people living with dementia (PLWD) in their own homes. Our main goal is to show the applicability of the entropy measures to time-series activity data analysis and to use the extracted measures as new engineered features that can be fed into inference and analysis models. The results of our experiments show that in most cases the combination of these measures can indicate the occurrence of healthcare-related events. We also find that different participants with the same events may have different measures based on one entropy measure. So using a combination of these measures in an inference model will be more effective than any of the single measures.","problem":"In-home IoT sensors produce long activity sequences, but raw activity alone does not explicitly describe routine, uncertainty, or changes in transitions between rooms. The paper studies interpretable entropy features for identifying changes that may coincide with healthcare-related events in people living with dementia.","contributions":["Application of Shannon entropy, first-order Markov-chain entropy rate, and neural entropy-production estimation to in-home activity data.","Evaluation on 9,370 person-days collected by a privacy-aware dementia-care monitoring platform.","Separate analysis of daytime and nighttime activity changes in relation to clinically recorded events."],"key_results":["In most evaluated cases, a combination of entropy measures indicated healthcare-related events more effectively than any single measure.","Participants experiencing the same event could show changes in different entropy measures, supporting a multi-feature inference approach."],"topics":["Activity Pattern Monitoring","Entropy Measures","Internet of Things","Dementia Care","In-Home Sensing","Health Monitoring","Time-Series Feature Engineering"],"bibtex":"@inproceedings{huang2022entropy,\n  author    = {Yushan Huang and Yuchen Zhao and Hamed Haddadi and Payam Barnaghi},\n  title     = {Using Entropy Measures for Monitoring the Evolution of Activity Patterns},\n  booktitle = {2022 IEEE 8th World Forum on Internet of Things (WF-IoT)},\n  year      = {2022},\n  pages     = {1--6},\n  doi       = {10.1109/WF-IoT54382.2022.10152050}\n}"},{"id":"cnc-wavelet-neural-network","title":"Thermal error modeling and analysis of CNC machine tools based on wavelet neural network","short_name":null,"former_titles":[],"authors":["Yushan Huang","Yu Chen","Zhenyu Hu"],"year":2021,"publication_date":"2021-01-15","venue":{"full":"2021 IEEE International Conference on Consumer Electronics and Computer Engineering","short":"ICCECE 2021"},"type":"conference","publisher":"IEEE","volume":null,"issue":null,"pages":"454-457","doi":"10.1109/ICCECE51280.2021.9342155","arxiv":null,"publisher_url":"https://ieeexplore.ieee.org/document/9342155","pdf_url":null,"code_url":null,"abstract":"Thermal error of CNC machine tools has become an important factor affecting its accuracy and it is of great significance to establish a thermal error prediction model with high prediction accuracy and good robustness. At present, the main way to reduce the thermal error is software compensation, which consists of the model of temperature and thermal error, and the model of temperature and thermal error curve's coefficients. However, most of the current models of temperature and coefficients only consider the relationship between nut temperature and thermal error, ignoring the bearing temperature, ambient temperature and speed. In addition, most of the models only consider the slope but ignore the intercept. Above all, the prediction accuracy of the present models is affected. In view of this problem, a prediction model of Z-axis thermal deformation is proposed, using nut temperature, ambient temperature, bearing temperature and speed to establish the relationship between them and both intercept and slope. First, a thermal error prediction model is established by wavelet neural network according to the slope and intercept of thermal error curve and bearing temperature, ambient temperature and speed. Finally, the prediction effect of the model is verified. The result shows that the prediction accuracy of this method is 97.1%, which greatly improves the prediction accuracy of thermal error and has good engineering practice significance.","problem":"Thermal deformation is a major source of CNC machine-tool positioning error. The work investigates data-driven thermal-error modeling with a wavelet neural network so that nonlinear temperature-to-error relationships can be analyzed for later compensation.","contributions":["A wavelet-neural-network formulation for modeling CNC machine-tool thermal error.","Analysis of the relationship between measured thermal behavior and machining error.","Experimental evaluation of the resulting prediction model."],"key_results":[],"topics":["CNC Machine Tools","Thermal Error Modeling","Wavelet Neural Networks","Precision Manufacturing","Error Compensation"],"bibtex":"@inproceedings{huang2021wavelet,\n  author    = {Yushan Huang and Yu Chen and Zhenyu Hu},\n  title     = {Thermal error modeling and analysis of {CNC} machine tools based on wavelet neural network},\n  booktitle = {2021 IEEE International Conference on Consumer Electronics and Computer Engineering (ICCECE)},\n  year      = {2021},\n  pages     = {454--457},\n  doi       = {10.1109/ICCECE51280.2021.9342155}\n}"},{"id":"feed-axis-lstm-thermal-error","title":"Thermal error modeling for machine tool feed axis based on LSTM neural network considering electro-control data","short_name":null,"former_titles":["Thermal error modeling of feed axis by LSTM neural network considering electric control data"],"authors":["Yushan Huang","Jihong Chen","Yu Chen","Guangda Xu"],"year":2021,"publication_date":"2021-10","venue":{"full":"Modern Manufacturing Engineering","short":"Modern Manufacturing Engineering"},"type":"journal","publisher":null,"volume":null,"issue":"10","pages":"25-32","doi":"10.16731/j.cnki.1671-3133.2021.10.004","arxiv":null,"publisher_url":"https://kns.cnki.net/kcms/detail/detail.aspx?doi=10.16731/j.cnki.1671-3133.2021.10.004","pdf_url":null,"code_url":null,"abstract":null,"problem":"Feed-axis thermal error evolves over time and depends on operating conditions recorded by the CNC controller. The paper addresses models that overlook this temporal history or exclude readily available electro-control signals, limiting accuracy and robustness across working conditions.","contributions":["An LSTM thermal-error model that represents the time dependence of feed-axis behavior.","Integration of temperature changes with electro-control data describing machine operating conditions.","Experimental comparison intended to test prediction accuracy and robustness under changing conditions."],"key_results":["The reported experiments found improved prediction accuracy and robustness when temperature variation and electro-control data were modeled together; the available verified sources did not expose the paper's numerical values."],"topics":["CNC Machine Tools","Feed-Axis Thermal Error","LSTM Networks","Electro-Control Data","Time-Series Modeling","Precision Manufacturing"],"bibtex":"@article{huang2021lstm,\n  author  = {Yushan Huang and Jihong Chen and Yu Chen and Guangda Xu},\n  title   = {Thermal error modeling for machine tool feed axis based on {LSTM} neural network considering electro-control data},\n  journal = {Modern Manufacturing Engineering},\n  number  = {10},\n  pages   = {25--32},\n  year    = {2021},\n  doi     = {10.16731/j.cnki.1671-3133.2021.10.004}\n}"},{"id":"vehicle-network-cloud-wsn","title":"Research on Vehicle Network Architecture and Key Technologies Based on Cloud Computing and Wsn","short_name":null,"former_titles":[],"authors":["Zikun Quan","Ning Gao","Yushan Huang"],"year":2019,"publication_date":"2019","venue":{"full":"Proceedings of the 2019 International Conference on Math and Engineering","short":"ICME 2019"},"type":"conference","publisher":"Francis Academic Press","volume":null,"issue":null,"pages":"574-577","doi":"10.25236/icme.2019.107","arxiv":null,"publisher_url":"https://doi.org/10.25236/icme.2019.107","pdf_url":"https://webofproceedings.org/proceedings_series/ESR/ICME%202019/D107.pdf","code_url":null,"abstract":"The article first briefly introduces the architecture and application direction of the car networking system. Then, based on the actual situation, from the two aspects of cloud computing and WSN, the discussion on the establishment of the Internet of Vehicles includes collecting, processing and applying traffic data. The aim is to help technical staff make use of cloud computing and WSN in vehicle networking systems and support progress in transportation systems.","problem":"Internet-of-Vehicles systems need an architecture that connects traffic-data collection, processing, and application across vehicles and road infrastructure. This paper discusses how cloud computing and wireless sensor networks can support those functions in an integrated transportation system.","contributions":["A review of vehicle-network architecture and its principal application directions.","A three-part architecture comprising a traffic cloud platform, intelligent vehicle system, and intelligent roadside system.","Discussion of cloud-computing and WSN roles in traffic-data collection, processing, and operational applications."],"key_results":[],"topics":["Internet of Vehicles","Cloud Computing","Wireless Sensor Networks","Intelligent Transportation Systems","Traffic Data","Vehicular Networks"],"bibtex":"@inproceedings{quan2019vehicle,\n  author    = {Zikun Quan and Ning Gao and Yushan Huang},\n  title     = {Research on Vehicle Network Architecture and Key Technologies Based on Cloud Computing and {Wsn}},\n  booktitle = {Proceedings of the 2019 International Conference on Math and Engineering},\n  year      = {2019},\n  pages     = {574--577},\n  doi       = {10.25236/icme.2019.107}\n}"}]