Uneven data coverage is a challenge in location-based crowd sensing. This paper investigates whether gamification can improve data coverage by motivating users to explore less frequently visited areas. Building on concepts of crowd sensing and gamification in the context of promoting changes in human mobility behavior, an online survey was developed. In the ongoing survey (n = 346), participants are presented with route-selection scenarios that vary in travel context and route length. Preliminary insights indicate that user behavior is influenced by a combination of contextual and perceptual factors, while basic gamification components may offer limited leverage for directing exploration. The findings suggest that more adaptive and personalized incentive strategies are needed to achieve a balanced and comprehensive coverage of urban environments in mobility data collection.

Deep neural networks are increasingly deployed on resource-constrained devices in IoT and pervasive computing environments, making real-time performance and heterogeneity of computing resources key challenges. Early exit networks offer a way to balance inference time and accuracy dynamically based on the input sample, but their use in dense prediction tasks like semantic segmentation has been limited and mainly focused on reducing FLOPs rather than materialized speed-ups. Inspired by input-sparse convolutions used in video processing, we propose a dynamic block-skipping approach for use in segmentation networks with early exits. This work presents a proof of concept to exploit sub-image level sparsity based on confidence thresholds to selectively bypass computations and enable faster inference. Our experiments confirm that the approach reduces runtime compared to dense convolution baselines aiding the deployment of segmentation networks in real-time, resource-constrained settings.

The demand for low latency, bandwidth efficiency, and privacy has driven distributed applications to the network edge, where heterogeneous, and untrusted, and uncontrolled networks pose challenges. This paper presents a software-defined overlay networking (SDON) middleware that simplifies application development through centralized control of edge resources while addressing many challenges prevalent at the edge. SDON enables applications to specify high-level requirements, such as service placement and communication goals. These are automatically translated into device-specific configurations and deployed on appropriate edge resources. The middleware is implemented as fully functional software, published as open-source, and will be evaluated in multiple edge computing use cases, to demonstrate its application optimization potential.

Production issues at Volkswagen in 2016 led to dramatic losses in sales of up to 400 million Euros per week. This example shows the huge financial impact of a working production facility for companies. Especially in the data-driven domains of Industry 4.0 and Industrial IoT with intelligent, connected machines, a conventional, static maintenance schedule seems to be old-fashioned. In this paper, we present an overview of the current state of the art in predictive maintenance for Industry 4.0. Based on a structured literature survey, we present a classification of predictive maintenance in the context of Industry 4.0 based on 249 publications. Additionally, we discuss identified challenges, i.e., complexity issues, as well as missing benchmark datasets that are relevant for production and the integration of machine learning.

The need for low latency, bandwidth efficiency, and privacy has driven the deployment of distributed applications to the network edge. However, edge environments introduce concrete challenges such as limited infrastructure control, constrained connectivity due to NAT or firewalls, and the heterogeneity of devices and network conditions. This paper introduces a software-defined overlay networking (SDON) middleware that addresses these issues by simplifying the development and deployment of edge applications through centralized control and dynamic overlay management. SDON allows applications to define high-level requirements, such as node and link characteristics and the network topology. These requirements are translated into device-specific configurations and enforced across suitable edge devices. We implemented our SDON middleware as a fully functional software and evaluated it in two edge computing use cases: i) routing for video streaming across middleboxed edge devices and ii) computation offloading on heterogeneous edge devices. Our results show that deployments via SDON, with centrally enforced optimizations, improve application performance by reducing mean streaming latency by 20 % and computation times by 22 %.

Process mining traditionally assumes centralized event data collection and analysis. However, modern Industrial Internet of Things systems increasingly operate over distributed, resource-constrained edge-cloud infrastructures. This paper proposes a structured approach for decentralizing process mining by enabling event data to be mined directly within the IoT systems edge-cloud continuum. We introduce ContinuumConductor a layered decision framework that guides when to perform process mining tasks such as preprocessing, correlation, and discovery centrally or decentrally. Thus, enabling privacy, responsive and resource-efficient process mining. For each step in the process mining pipeline, we analyze the trade-offs of decentralization versus centralization across these layers and propose decision criteria. We demonstrate ContinuumConductor at a real-world use-case of process optimazition in inland ports. Our contributions lay the foundation for computing-aware process mining in cyber-physical and IIoT systems.

3D printing has revolutionized DIY (Do-It-Yourself) IoT prototyping, enabling cost-effective, creative custom device creation. However, this freedom also presents challenges due to the interplay between components within an IoT design, which can influence the overall utility and performance of the prototype. Optimizing these designs is difficult due to limited means of estimating their efficacy. To address this, we introduce TOAD, a novel tool for profiling IoT prototypes and gauging their performance impact. TOAD uses thermal imaging and video analysis to extract and compare design performance characteristics. Unlike existing solutions that only profile overall performance, our tool assesses component interactions and overall design effects. It offers an affordable, non-intrusive method without needing device access or code instrumentation. Extensive benchmarks show TOAD accurately extracts performance data, aiding in selecting the best design for IoT applications. Additionally, it provides insights into how casing factors like thickness and material influence thermal behavior and performance. We demonstrate practical applications by optimizing offloading decisions based on thermal behavior, highlighting casing impacts on design performance. TOAD paves the way for ef-ficient IoT prototype designs, offering a better understanding of component interactions and significantly enhancing the utility of custom IoT designs and their effectiveness.

WebAssembly is a portable binary instruction format designed to serve as a compilation target for high-level languages. While originally developed to run performance-intensive applications directly in Web browsers, WebAssembly supports these days a number of different hardware platforms across the compute continuum. This makes it a promising option to run services for training and inference in Federated Learning. To the best of our knowledge, there have been only a few practical approaches to realize Federated Learning usingWebAssembly. Therefore, in this paper, we present a framework to achieve this. Our prototypical implementation shows that WebAssembly-based Federated Learning applications are highly portable while providing acceptable runtime overhead during model training.

The number of people living in cities is growing every year. This leads to challenges such as scarcity of natural resources, demographic change, and ongoing urbanization, to which smart cities promise a solution. With the help of the growing Internet of Things, these visions could become reality. Already there are many sensors deployed in local households but collected data streams are often used only locally. One of the main concerns of sharing collected data with external services and applications is the citizen's privacy. Citizens only want to share data that is truly necessary for the service, and not all data they collect. This work presents an approach that reduces data collection to the required minimum within citizen centric sensor network. Therefore services requesting citizens' data define preprocessing steps that are executed locally at the citizens' devices. Preprocessed data is then sent to the service, reducing network load and increasing citizens' privacy by only sending relevant information. A major challenge is to describe these preprocessing steps in a comprehensible manner. This ensures that they can be understood by average citizens enabling them to make informed decisions. In particular, it must be ensured that only data specified within the steps is sent to the service. To evaluate our approach regarding comprehensibility and usability a prototype has been implemented and used within a user study. The results show that for the majority of tasks, both users with and without computer science backgrounds were able to understand the given preprocessing step sequences. Furthermore, developers were able to define preprocessing step sequences themselves.

Recently, there is a growing trend where endusers personally engage in computationally-intensive applications, particularly in domains such as AI, AR/VR, and simulation. Consequently, there arises an increasing demand to enhance the capabilities of local devices by offloading computational work to remote resources, which can also be provided by endusers. To address the requirements of modern peer-to-peer (P2P) offloading architectures, we have developed Wasimoff—an innovative browser-based framework designed for dynamic volunteer computing. Leveraging WebAssembly, applications can securely run in an isolated browser environment. Device owners can effortlessly share their available CPU cycles and participate in the collaborative network by accessing a designated URL in their browser. To assess Wasimoff’s performance, we conducted a real-world evaluation with devices located in both Germany and the USA. The results affirm Wasimoff’s ability to scale linearly as the system’s entities increase. Moreover, Wasimoff consistently maintains execution speed even under congested system conditions.

Edge artificial intelligence (AI) is an innovative computing paradigm that aims to shift the training and inference of machine learning models to the edge of the network. This paradigm offers the opportunity to significantly impact our everyday lives with new services such as autonomous driving and ubiquitous personalized health care. Nevertheless, bringing intelligence to the edge involves several major challenges, which include the need to constrain model architecture designs, the secure distribution and execution of the trained models, and the substantial network load required to distribute the models and data collected for training. In this article, we highlight key aspects in the development of edge AI in the past and connect them to current challenges. This article aims to identify research opportunities for edge AI, relevant to bring together the research in the fields of artificial intelligence and edge computing.

In the edge-cloud continuum, computational resources may be available in the cloud and at the edge of the network. If services are deployed on these resources in a distributed fashion, the latency between the participating nodes may become a critical factor. This is especially the case if resources from different clusters are used and runtime dependencies between services exist. In addition, edge-cloud landscapes are volatile, i.e., both the topology as well as Quality of Service parameters like latency may change over time. In this paper, we present a service placement strategy suited to deploy services in the cloud-edge continuum, aiming at reducing latency. Our placement strategy takes into account the dynamic nature of the edge-cloud continuum. In addition, dependencies between services are discovered at runtime. To allow this, we present a framework extending Kubernetes, which identifies dependencies automatically. We evaluate our approach and can show that we are able to reduce the latency significantly.

Advanced AI applications have become increasingly available to a broad audience, e.g., as centrally managed large language models (LLMs). Such centralization is both a risk and a performance bottleneck - Edge AI promises to be a solution to these problems. However, its decentralized approach raises additional challenges regarding security and safety. In this paper, we argue that both of these aspects are critical for Edge AI, and even more so, their integration. Concretely, we survey security and safety threats, summarize existing countermeasures, and collect open challenges as a call for more research in this area.

Affective systems are a class of pervasive systems that aim to support humans at all levels, including critical interactions with medical services or learning platforms by measuring, understanding, and acting on behalf of affect, feelings, and emotions. In order to react appropriately, affective systems need to correctly read their human counterpart. In this paper, eye tracking is proposed as a non-invasive tool to measure the affective quality of videos. We collect gaze data from 175 subjects watching eight videos and assess how fixation features react to the affective content. While several features are biased by video-specific effects of dynamics and sound and show no sensitivity to affect after correcting for these effects, the average fixation duration and dwell time preserve discriminating patterns. On videos triggering negative affect, the gaze is focused, with meticulous examinations of the risk inducing element - characterized by longer dwells and shorter fixation duration. Positively connoted videos lead to comprehensive examinations of the entire scene, marked by shorter dwells and fixation duration.

In the evolving landscape of distributed computing frameworks, Wasimoff emerges as an innovative middleware with a browser-based execution environment designed to facilitate computation offloading to heterogeneous volunteer devices. By leveraging WebAssembly as an executable format, Wasimoff allows users to seamlessly contribute computational resources through a simple web interface, thereby eliminating complex setup requirements. This demonstration presents the zero-setup functionality of the Wasimoff provider, and showcases the ease of connecting personal devices to a dynamic resource pool, highlighting the scaling efficiency and multitasking capabilities of the framework. By enabling spontaneous, barrier-free participation, Wasimoff transforms idle device capacity into a collaborative computational resource.

Real-time simulation in augmented reality environments is an important area of research. The simulation model used in this paper simulates the musculoskeletal system, which has many potential applications in physiotherapy, medical education, and rehabilitation. To enable such real-time simulations in augmented reality, high-performance computing clusters are typically required, which can be costly and infeasible in terms of high network delay. This paper proposes SimEdge, a domain-specific edge computing system that enables real-time pervasive simulation in augmented reality environments on heterogeneous providers. This is done by smart offloading decisions based on a continuously updated list of resource providers. SimEdge builds on previous work using surrogate models and leverages scheduling improvements and context-aware data and task placement. Significant improvements in frame rate and input lag are achieved, as well as a reduction in energy consumption. The experimental results show that the proposed approach leads to an 8.3-fold increase in frame rate and a 65.14% reduction in input lag, as well as a 35.07% reduction in energy consumption compared to the local baseline system. These results demonstrate the effectiveness of the approach in improving efficiency and responsiveness.

When the demand for computationally-intensive applications exceeds the capabilities of a single device, computation offloading can help to leverage remote resources. Currently, most systems only use cloud and grid resources as offloading targets, ignoring the vast amount of end-user devices that might contribute with their idle CPU cycles. Even though these edge devices have great potential, two main reasons make their usage more complicated: First, the devices are highly heterogeneous and unreliable. Second, numerous middleboxes on the Internet impose barriers, making communication between end-user devices difficult. Thus, in this paper, we overcome these two obstacles and propose an offloading system composed not only of cloud and grid resources but also of geographically distributed and heterogeneous end-user devices. We deploy this system in realistic environments and evaluate its performance under real-world conditions. Our results indicate that offloading to edge devices owned by end users can compete with cloud and grid offloading while adding minimal communication overhead. Our empirical findings support the hypothesis that edge computing can be a cost-efficient alternative to traditional offloading systems.