3 research outputs found
Smart Semiconductor Testing Systems: Fusion of Embedded AI And Scalable Data Pipelines
No full textSemiconductor companies are challenged by increasingly complex testing requirements coming from their customers and technology, as the design of modern System-on-Chips (SoCs) evolves into multi-chiplets. New AI-driven paradigms are needed to facilitate a massively parallel high-throughput test methodology while still allowing tight test channel characterization that affects both yield and performance of the complex designs. With the increase in chip complexity and design autonomy of third-party chiplets, a whole new add-on market for in-die and package-level testing is created, and access to the test systems is kept tightly.
A new architecture for the system under test (SUT) is presented based on decisions at test time and embedded intelligence combined with a distributed AI-based device that abstracts the test flow towards a Domain-Specific Language (DSL) API. This new approach is complemented by a novel design-to-test procedure and scalable machine learning pipelines on chiplet level. With this approach, a Semantic Web-based ecosystem of tools and libraries is created that links simulators and correlators and allows engineers to compose powerful packages of tasks, lab experiments, and production data mining.
Traditional semiconductor integrated circuit (IC) test systems are fast reaching their limits with respect to both test data throughput in the order of petabytes and complexity of platform and device under test which need to be test parallelized in order to ensure operational use. For SoCs and their Subsystems, an architecture and implementation of a non-standard test methodology is proposed that is distributed, massively parallel, and AI driven. The ambition is to merge the fabrication test domain with various application domains in order to perform heterogeneous tests
Emotional Intelligence in Artificial Agents: Leveraging Deep Multimodal Big Data for Contextual Social Interaction and Adaptive Behavioral Modelling
No full textAs artificial agents develop beyond mere tools and begin to perform roles traditionally associated with humans, expectations of their performance are equally evolving. Not only must agents be able to accomplish their tasks; but they must also be able to do so in a manner that observers would consider socially or contextually appropriate. For social interaction where the agent and human are co-performers, adherence to social cues that signal emergent aspects of a relationship such as intimacy or status is paramount to the experience of the interacting humans. For autonomous agents who function alone, adaptive behavioral modeling and user state awareness are critical to the impact of the agent’s actions on humans. Such contextual social behavior is a requirement for complex applications including physically located social robots, virtual avatars emerging in gaming, online social environments, or customer service interactions, and proactive virtual assistants. Humans have sophisticated socio-emotional capacities that enable them to behaviorally coordinate their interactions with others, inferring mental states that may lie far beyond explicit observable cues. Furthermore, emotional expressions are multimodal and are the result of a complex interaction between inherent affective states and contextual interaction. The Human Centered Intelligent Systems conceptual framework describes a pathway whereby artificial agents may also achieve aspects of this intelligence through rich user state modeling based on deep multimodal analysis of big data that can capture the social behavior and interaction context. In this chapter, we describe this "user-state" modeling approach and exemplify its applicability to a spectrum of agent applications
Optimizing Edge Computing for Big Data Processing in Smart Cities
No full textThe surge of big data and IoT in smart cities requires effective computational models to process massive amounts of real-time data. Edge computing emerges as an innovative solution by minimizing latency, improving security, and maximizing energy efficiency. This paper investigates the convergence of AI-based edge computing for big data processing through a study of four sophisticated algorithms: Federated Learning, TinyML, Edge-Optimized CNNs, and Adaptive Data Compression. Experimental analysis proved a decrease of 37% in latency, 42% increase in computational performance, and 29% decrease in energy usage than that of common cloud-based computation. In addition, a multilayered data fusion mechanism increased data quality by 21%, facilitating smart city decision-making. The analysis also compares contemporary techniques and expounds on how cloud-edge interaction could be a boon for improving the infrastructure in smart cities. Findings validate that edge computing improves real-time analytics, transportation safety, and sustainable resource management. Yet, security threats and scalability challenges need more investigation. Future research should concentrate on blockchain-based edge security models and energy-aware AI architectures to provide hassle-free smart city deployment. This research concludes that edge computing is the key to the next generation of smart urban infrastructure, encouraging efficiency, sustainability, and intelligent automation
