Condition-awareness regarding electrical and electronic components is not only significant for predictive maintenance of automotive vehicles but also plays a crucial role in ensuring the operational safety by supporting the detection of anomalies, faults, and degradations over lifetime. In this paper, we present a novel control strategy that combines stochastic dynamic control method with condition-awareness for safe automated driving. In particular, the effectiveness of condition-awareness is supported by two distinct condition-monitoring functions. The first function involves the monitoring of a vehicle's internal health condition using model-based approaches. The second function involves the monitoring of a vehicle's external surrounding conditions, using machine learning and artificial intelligence approaches. For the quantification of current conditions, the results from these monitoring functions are used to create system health indices, which are then utilized by a safety control function for dynamic behavior regulation. The design of this safety control function is based on a chance-constrained model predictive control model, combined with a control barrier function for ensuring safe operation. The novelty of the proposed method lies in a systematic integration of monitored external and internal conditions, estimated component degradation, and remaining useful life, with the controller's dynamic responsiveness. The efficacy of the proposed strategy is evaluated with adaptive cruise control in the presence of various sensory uncertainties.

With the advance of Artificial Intelligence, Deep Neural Networks are widely employed in various sensor-based systems to analyze operational conditions. However, due to the inherently nondeterministic and probabilistic natures of neural networks, the assurance of overall system performance could become a challenging task. In particular, soft errors could weaken the robustness of such networks and thereby threaten the system's safety. Conventional fault-tolerant techniques by means of hardware redundancy and software correction mechanisms often involve a tricky trade-off between effectiveness and scalability in addressing the extensive design space of Deep Neural Networks. In this work, we propose a Reinforcement-Learning-based approach to protect neural networks from soft errors by addressing and identifying the vulnerable bits. The approach consists of three key steps: (1) analyzing layer-wise resiliency of Deep Neural Networks by a fault injection simulation; (2) generating layer-wise bit masks by a Reinforcement-Learning-based agent to reveal the vulnerable bits and to protect against them; and (3) synthesizing and deploying bit masks across the network with guaranteed operation efficiency by adopting transfer learning. As a case study, we select several existing neural networks to test and validate the design. The performance of the proposed approach is compared with the performance of other baseline methods, including Hamming code and the Most Significant Bits protection schemes. The results indicate that the proposed method exhibits a significant improvement. Specifically, we observe that the proposed method achieves a significant performance gain of at least 10% to 15% over on the test network. The results indicate that the proposed method dynamically and efficiently protects the vulnerable bits compared with the baseline methods.

With globalization and outsourcing trends, modern industrial companies often rely on an extensive network of suppliers to construct sophisticated products. To maintain effective production planning and scheduling, integrating and managing the extensive information from supply chain has become increasingly critical. In particular, industrial companies, particularly those aiming to achieve Industry 4.0, enable to collect and analyze data related to their supply chains. Due to the vast amount of collected data, there is a continuous challenge in integrating and analyzing dependencies within the supplier network. While the development of Artificial Intelligence (AI) offers a promising solution for extracting and analyzing features from data, the inherently opaque and training-intensive nature of AI-enabled methods still present obstacles to effectively and efficiently analyzing information. To cope with this issue, this paper presents a knowledge-enhanced framework to support supply chain information integration and analysis by combining Knowledge Base (KB) and Graph Neural Networks (GNN). Specifically, constructing a KB enables the integration of extensive collected data with domain knowledge to generate structured and relational information. These knowledge-enhanced data support the training of GNN to encode information about supply chains. The resulting embeddings enable multiple inference tasks for analyzing graph-based data, supporting supply chain management. The case studies cover the usage of encoded embeddings for node classification, link prediction, and scenario classification. The proposed GNN outperforms baseline methods, demonstrating a promising solution for analyzing graph-based data in the context of supply chain management.

Prognostics and Health Management (PHM) services play a critical role in maintaining system performance by monitoring and managing the status of industrial Cyber-Physical Systems (CPS). These services typically involve analyzing multi-dimensional data from various sensors to extract spatial and temporal features. The advancement of data-driven methods, such as Machine Learning (ML) and Deep Learning (DL), provides an end-to-end solution for straightforwardly manipulating sensory data and learning patterns. In addition to these methods, the rise of Large Language Models (LLM) also facilitates the analysis and comprehension of multi-dimensional data. Compared to ML/DL models, LLM demonstrate powerful capabilities due to their extensive training data and cross-domain knowledge. However, continuous challenges remain in seamlessly detecting anomalies by integrating domain knowledge with these methods to effectively and efficiently monitor system status for health management (e.g., fault propagation via causality analysis). To alleviate this issue, this paper presents a functional framework to support the health management in CPS by leveraging the integration of LLM and domain knowledge. As case studies, this work uses the Tennessee Eastman (TE) process and an Ultra-Processed (UP) food manufacturing plant to evaluate the performance of integrating LLM with domain knowledge by identifying and analyzing the causality of collected data across sensors. Compared to baseline methods that rely solely on the LLM, the results show an increase in F1-score from 0.15 to 0.62 for the TE process, and from 0.05 to 0.64 for the UP process.

In this chapter, we explore how new technologies and requirements affect current design methodologies for mechatronics. We investigate gaps and directions needed for the methodologies of tomorrow in view of trends affecting mechatronics and current state of the art. To fully reap the opportunities of mechatronics with advances in materials, sensors, additive manufacturing, AI, computing and communication, but also to handle new requirements and regulations, there is a need for new methodologies and architectures. We introduce the concept of “MechaOps” and related considerations that promise to assist in enhancing scalability, smartness, performance and sustainability for extended mechatronic products that collaborate with a smart infrastructure, humans and other mechatronic systems. MechaOps refers to the integration of the concepts of Mechatronics and DevOps. As opposed to DevOps in software engineering, MechaOps encompasses data gathering, upgrades/downgrades as well as reconfigurations considering both mechanics and/or software in a mechatronic product. With the life-cycle view implied by the MechaOps concept, it becomes essential to design for upgrading, downgrading, maintenance, reuse and refurbishment. The development of new methodologies requires overcoming disciplinary gaps, with specific considerations of novel architectures including digital twins, interactions with humans, other systems and a smart infrastructure, the role of AI in mechatronics, and in assuring trustworthiness and sustainability. We believe that new methodologies and architectures will initially be especially relevant for high-end systems, supporting the creation of adaptable and flexible mechatronics products and services with improved performance and reduced environmental footprint.

This study demonstrates a data-driven decision support system to aid in rectification of prosthetic sockets aimed at improving overall comfort perceived by amputees. Prosthetic technology, particularly in the realm of socket design, plays a pivotal role in rehabilitation for individuals with limb amputations. Prosthetic sockets, which serve as the critical interface between the residual limb and the artificial limb, enable amputees to walk without the need for invasive implants that connect directly to the bone of the residual limb. This study focuses on the role of intra-socket pressure in socket performance and its impact on optimal socket rectifications for improving comfort in transfemoral amputees. Employing thin Force Sensing Resistor (FSR) sensors, the research measures dynamic pressure variations across individual gait cycles. To explore the effects of altered pressure distribution on socket performance, a clinical trial was conducted consisting of four different socket configurations across several participants, one of which was with no pad inserted and three of which incorporated a silicone pad to modify the dynamic pressure profiles. With data from multiple participants including specific dynamic pressure features extracted from FSR sensors, and subjective feedback of comfort, a Multi-Layer Perceptron (MLP) model is trained to establish predictive relationships between intra-socket pressure and appropriate rectification action. The findings suggest that the MLP agent is more accurate at suggesting rectification actions to prosthetists when compared to simpler classification algorithms such as Random Forest, XGBoost and Logistic regression, laying the foundation for future advancements in prosthetic design.

This paper presents a reconfigurable near-sensor anomaly detection processor to real-time monitor the potential anomalous behaviors of amputees with limb prostheses. The processor is low-power, low-latency, and suitable for equipment on the prostheses and comprises a reconfigurable Variational Autoencoder (VAE), a scalable Self-Organizing Map (SOM) Array, and a window-size-adjustable Markov Chain, which can implement an integrated miniaturized anomaly detection system. With the reconfigurable VAE, the proposed processor can support up to 64 sensor sampling channels programmable by global configuration, which can meet the anomaly detection requirements in different scenarios. A scalable SOM array allows for the selection of different sizes based on the complexity of the data. Unlike traditional time accumulation-based anomaly detection methods, the Markov Chain is utilized to detect time-series-based anomalous data. The processor is designed and fabricated in a UMC 40-nm LP technology with a core area of 1.49 mm2 and a power consumption of 1.81 mW. It achieves real-time detection performance with 0.933 average F1 Score for the FSP dataset within 24.22 s, and 0.956 average F1 Score for the SFDLA-12 dataset within 30.48 s, respectively. The energy dissipation of detection for each input feature is 43.84 nJ with the FSP dataset, and 55.17 nJ with the SFDLA-12 dataset. Compared with ARM Cortex-M4 and ARM Cortex-M33 microcontrollers, the processor achieves energy and area efficiency improvements ranging from 257×, 193× and 11×, 8×, respectively. IEEE

Human Activity Recognition (HAR) refers to a field that aims to identify human activitiesby adopting multiple techniques. In this field, different applications, such as smart homes andassistive robots, are introduced to support individuals in their Activities of Daily Living (ADL)by analyzing data collected from various sensors. Apart from wearable sensors, the adoption ofcamera frames to analyze and classify ADL has emerged as a promising trend for achieving theidentification and classification of ADL. To accomplish this, the existing approaches typically rely onobject classification with pose estimation using the image frames collected from cameras. Given theexistence of inherent correlations between human–object interactions and ADL, further efforts areoften needed to leverage these correlations for more effective and well justified decisions. To this end,this work proposes a framework where Graph Neural Networks (GNN) are adopted to explicitlyanalyze human–object interactions for more effectively recognizing daily activities. By automaticallyencoding the correlations among various interactions detected through some collected relational data,the framework infers the existence of different activities alongside their corresponding environmentalobjects. As a case study, we use the Toyota Smart Home dataset to evaluate the proposed framework.Compared with conventional feed-forward neural networks, the results demonstrate significantlysuperior performance in identifying ADL, allowing for the classification of different daily activitieswith an accuracy of 0.88. Furthermore, the incorporation of encoded information from relational dataenhances object-inference performance compared to the GNN without joint prediction, increasingaccuracy from 0.71 to 0.77. 

Gait analysis helps clinical assessment and achieves comfortable prosthetic designs for lower limb amputees, in which accurate gait phase recognition is a key component. However, gait phase detection remains a challenge due to the individual nature of prosthetic sockets and limbs. For the first time, we present a gait phase recognition approach for transfemoral amputees based on intra-socket pressure measurement. We proposed a multiple GMM-HMM (Hidden Markov Model with Gaussian Mixture Model emissions) method to label the gait events during walking. For each of the gait phases in the gait cycle, a separate GMM-HMM model is trained from the collected pressure data. We use gait phase recognition accuracy as a primary metric. The evaluation of six human subjects during walking shows a high accuracy of over 99% for single-subject, around 97.4% for multiple-subject, and up to 84.5% for unseen-subject scenarios. We compare our approach with the widely used CHMM (Continuous HMM) and LSTM (Long Short-term Memory) based methods, demonstrating better recognition accuracy performance across all scenarios.