Despite its enormous economical and societal impact, lack of human-perceived control and safety is re-defining the design and development of emerging AI-based technologies. New regulatory requirements mandate increased human control and oversight of AI, transforming the development practices and responsibilities of individuals interacting with AI. In this paper, we present the SPATIAL architecture, a system that augments modern applications with capabilities to gauge and monitor trustworthy properties of AI inference capabilities. To design SPATIAL, we first explore the evolution of modern system architectures and how AI components and pipelines are integrated. With this information, we then develop a proof-of- concept architecture that analyzes AI models in a human-in-the- loop manner. SPATIAL provides an AI dashboard for allowing individuals interacting with applications to obtain quantifiable insights about the AI decision process. This information is then used by human operators to comprehend possible issues that influence the performance of AI models and adjust or counter them. Through rigorous benchmarks and experiments in real- world industrial applications, we demonstrate that SPATIAL can easily augment modern applications with metrics to gauge and monitor trustworthiness, however, this in turn increases the complexity of developing and maintaining systems implementing AI. Our work highlights lessons learned and experiences from augmenting modern applications with mechanisms that support regulatory compliance of AI. In addition, we also present a road map of on-going challenges that require attention to achieve robust trustworthy analysis of AI and greater engagement of human oversight.

Littering is a significant environmental concern that causes significant damage to the natural ecosystem and contributes adversely to human health. Monitoring litter accumulation is currently labour-intensive and costly, often resulting in action being taken only once the environment has already become polluted. We contribute LIZARD, a novel pervasive sensing solution for detecting and monitoring plastics that is tailored to autonomous vehicles. LIZARD relies on an innovative sensing pipeline that combines thermal imaging and optical sensing. The intuition is to rely on thermal dissipation patterns to identify larger (macro) plastics and use optical sensing to sample area with the highest density of plastics to identify smaller (micro and meso) plastics. Ours is the first pervasive sensing solution that can detect microplastics in the environment and be integrated into autonomous vehicles. Indeed, state-of-the-art solutions are either limited to laboratory analysis with special instruments or rely on manual observation without being able to identify the smallest plastics – which often are the most dangerous. We evaluate LIZARD through rigorous experiments that combine controlled laboratory settings and in-the-field measurements carried out in three real-world locations to evaluate LIZARD. Our results show that LIZARD can be used to detect plastics of different sizes with an accuracy of up to 80%. The performance depends on the diameter of the plastics, the background surface, and the luminosity of the environment. We also demonstrate that our solution can be easily integrated with ground drones, enabling (semi-)autonomous litter monitoring. Our work offers an innovative way to harness pervasive sensing to address an important global (environmental) sustainability challenge while paving the way toward improved monitoring of the accumulation of harmful plastic fragments in the environment.

The inherent social characteristics of humans make them prone to adopting distributed and collaborative applications easily. Although fundamental methods and technologies have been defined and developed over the years to construct these applications, their adoption in practice is uncommon because end-users may be puzzled about how to use them without much hassle. Indeed, commonly, these applications require a certain level of technical expertise and awareness to use them correctly. Fortunately, AI-chatbot interventions are envisioned to assist and support various human tasks. In this paper, we contribute pervasive chatbots as a solution that fosters a more transparent and user-friendly interconnection of devices in distributed and collaborative environments. Through two rigorous user studies, firstly, we quantify the perception of users toward distributed and collaborative applications (N = 56 participants). Secondly, we analyze the benefits of adopting pervasive chatbots when compared with the chatbot reference model designed for assistance and recommendations (N = 24 participants). Our results suggest that pervasive chatbots can significantly enhance the practicability of distributed and collaborative applications, reducing the time and effort needed for collaboration with surrounding devices by 57%. With this information, we then provide design and development implications to integrate pervasive chatbot interventions in distributed and collaborative environments. Moreover, challenges and opportunities are also provided to highlight the remaining issues that need to be addressed to realize the full vision of pervasive chatbots for any multi-device application. Our work paves the way towards the proliferation of sophisticated and highly decentralized computing environments that are easily interconnected.

We demonstrate SPATIAL, a proof-of-concept system that augments modern applications with capabilities to analyze trustworthy properties of AI models. The practical analysis of trustworthy properties is key to guaranteeing the safety of users and overall society when interacting with AI -driven applications. SPATIAL implements AI dashboards to introduce human-in-the-loop capabilities for the construction of AI models. SPATIAL allows different stakeholders to obtain quantifiable insights that characterize the decision making process of AI. This information can then be used by the stakeholders to comprehend possible issues that influence the performance of AI models, such that the issues can be resolved by human operators. Through rigorous benchmarks and experiments in a real-world industrial application, we demonstrate that SPATIAL can easily augment modern applications with metrics to gauge and monitor trustworthiness. However, this, in turn, increases the complexity of developing and maintaining the systems implementing AI. Our work paves the way towards augmenting modern applications with trustworthy AI mechanisms and human oversight approaches.

We analyze the impact of poisoning attacks on autonomous drones and demonstrate how explainable artificial intelligence techniques can be employed to detect them. We then delve into the risks, opportunities, and research challenges, ultimately paving the way for city-scale deployments of autonomous drones.

We contribute PRINCE, an innovative sensing solution capable of accurately estimating the energy consumption of applications executing on a wider range of smart and IoT devices, including smartwatches, wearables and autonomous drones, without the need for direct instrumentation of the device. Modern devices lack detachable batteries or are sealed, making it challenging to profile their energy consumption. In this demo, we showcase PRINCE, a proof-of-concept prototype that provides precise energy consumption measurements of applications running in devices with a single (thermal) photo. PRINCE harnesses the thermal radiation (heat) generated by the processing units of the device, which is released through the device casing. This allows PRINCE to derive accurate energy estimations of application execution. Extensive benchmarks that compare PRINCE with traditional solutions, such as Monsoon power monitor, demonstrate that PRINCE provide similar performance levels but does not require any instrumentation, facilitating the profiling of the energy consumption of devices.

Modern system architectures are rapidly adopting AI-based functionality. As a result, new requirements about software trustworthiness must be considered during the entire software development life cycle of applications. While several requirement management tools are available to track and monitor requirements over time, it is still unknown to what extent these tools can cope with these new demands imposed by AI. In this paper, we contribute by performing a qualitative and quantitative analysis of different requirement management tools and their performance in managing AI-related requirements effectively. Through a rigorous analysis performed by a consortium formed by different industry and academic partners, we evaluate the suitability of five different requirement management tools. Our results indicate that while several tools are available for managing requirements, it is currently challenging to find a tool that can manage AI requirements mainly because tools do not comply with the required aspects imposed by regulatory entities. Lastly, we also shared our lessons learned and experiences from selecting requirement tools that can be used in team-based consortium projects.

We contribute by presenting a framework that re-purposes off-the-shelf and low-cost components into integrated solutions that are easy to scale and deploy in the wild. We demonstrate the applicability of our framework in the context of produce quality estimation to advance the digital transformation of existing agricultural practices. The deployment of off-the-shelf technologies is critical to foster its large-scale adoption and to accelerate the automation of human manual activities. Through rigorous experiments using our proposed framework, first we demonstrate that individual off-the-shelf light sensors (in three different spectra, green, red and blue) can be easily re-purposed for produce quality estimation, and that this monitoring solution can be further integrated into off-the-shelf nano-drones to support dynamic produce quality estimation at different altitudes without degrading its estimation performance. Our work paves the way towards practical guidelines that can be used to assemble complex off-the-shelf components in a plug and play fashion.

Federated learning (FL) is a promising privacy-preserving solution to build powerful AI models. In many FL scenarios, such as healthcare or smart city monitoring, the user’s devices may lack the required capabilities to collect suitable data, which limits their contributions to the global model. We contribute social-aware federated learning as a solution to boost the contributions of individuals by allowing outsourcing tasks to social connections. We identify key challenges and opportunities, and establish a research roadmap for the path forward. Through a user study with N = 30 participants, we study collaborative incentives for FL showing that social-aware collaborations can significantly boost the number of contributions to a global model provided that the right incentive structures are in place.

Hand-grip strength is a widely recognized indicator of muscle strength and overall health of individuals, particularly among older adults. Hand-grip strength measurements are typically obtained using dynamometers or specifically tailored devices, limiting the context in which measurements can be taken to health checks and clinical settings. In this demo, we showcase a new smart ring, namely HIPPO. The smart ring implements an innovative approach that offers a non-intrusive and opportunistic way to extract handgrip strength measurements from individuals. HIPPO re-purposes off-the-shelf light sensors available in existing wearable devices, e.g., smartwatches, and exploits the principle of light reflectivity, such that as an individual interacts with everyday objects, changes in their surfaces can be used to derive the hand-grip measurements.

<p>Autonomous drones are reaching a level of maturity when they can be deployed in cities to support tasks ranging from medicine or food delivery to environmental monitoring. These operations rely on powerful AI models integrated into the drones. Ensuring these models are robust is essential for operating in cities as any errors in the decisions of the autonomous drones can cause damage to the citizens or the urban infrastructure. We contribute a research vision for trustworthy city-scale deployments of autonomous drones. We highlight current key requirements and challenges that have to be fulfilled for achieving city-scale autonomous drone deployments. In addition, we also analyze the complexity of using XAI methods to monitor drone behavior. We demonstrate this by inducing changes in AI model behavior using data poisoning attacks. Our results demonstrate that XAI methods are sensitive enough to detect the possibility of a data attack, but a combination of multiple XAI methods is better to improve the robustness of the estimation. Our results also suggest that currently, the reaction time to counter an attack in city-scale deployment is large due to the complexity of the XAI analysis.</p>

Hand-grip strength is widely used to estimate muscle strength and it serves as a general indicator of the overall health of a person, particularly in aging adults. Hand-grip strength is typically estimated using dynamometers or specialized force resistant pressure sensors embedded onto objects. Both of these solutions require the user to interact with a dedicated measurement device which unnecessarily restricts the contexts where estimates are acquired. We contribute HIPPO, a novel non-intrusive and opportunistic method for estimating hand-grip strength from everyday interactions with objects. HIPPO re-purposes light sensors available in wearables (e.g., rings or gloves) to capture changes in light reflectivity when people interact with objects. This allows HIPPO to non-intrusively piggyback everyday interactions for health information without affecting the user's everyday routines. We present two prototypes integrating HIPPO, an early smart glove proof-of-concept, and a further optimized solution that uses sensors integrated onto a ring. We validate HIPPO through extensive experiments and compare HIPPO against three baselines, including a clinical dynamometer. Our results show that HIPPO operates robustly across a wide range of everyday objects, and participants. The force strength estimates correlate with estimates produced by pressure-based devices, and can also determine the correct hand grip strength category with up to 86% accuracy. Our findings also suggest that users prefer our approach to existing solutions as HIPPO blends the estimation with everyday interactions.

Pervasive technologies are supporting the digital transformation of agriculture practices. A major limitation of these solutions is that they are difficult to deploy at large-scale. Indeed, these solutions are typically built with specific designs and unique characteristics, making them difficult to adjust to different environments. Moreover, these solutions also necessitate the deployment of dedicated on-site infrastructure to operate them. A key way to overcome this problem is to rely on off-the-shelf components and their available technologies. Unfortunately, off-the-shelf components are not yet at the stage of easy plug and play integration, requiring specifications and guidelines on how to integrate them seamlessly. In this demonstration, we show the practical design and development of nano-drones for produce quality monitoring using off-the-shelf components. By incrementally re-purposing and integrating individual components, we demonstrate that it is possible to combine off-the-shelf technologies into an integrated single solution. Through rigorous experiments, we analyze the performance of our combined off-the-shelf solution and its effectiveness to perform produce quality monitoring. In addition, we also share lessons learned and experiences from building our plug and play off-the-shelf prototype.