Edge computing seeks to optimize service provision over enterprise-owned infrastructure near the end-user at the network’s edge. However, it misses out on the opportunity to utilize user-owned hardware at the extreme edge of the network as workers in a sharing economy. In this work, we build upon the existing Incentive Vacation Queueing (IVQ) model and develop the Virtual Kiosk Model (VKM) to analyze service availability and the dynamics of multiple workers’ participation in the provision of a service on the extreme edge. We formulate an optimization problem to minimize total cost of incentive payments while maintaining service availability under temporal constraints. We propose the Model-based Incentive Strategy at the Edge (MISE) algorithm to iteratively adjust incentives in real-time. MISE is compared against traditional numerical optimizers and a baseline naive approach that greedily focuses on minimizing incentives. Our findings demonstrate that MISE ensures sustained service availability without overburdening the workers at a cost acceptable to the service provider, striking a crucial balance in the management of extreme edge computing resources.

Extreme Edge Sensing (EES) offers an enhanced approach to efficient remote sensing by utilizing the computational capabilities of user devices for immediate data processing. In contrast to traditional Mobile Crowd Sensing (MCS), EES provides both data collection and local data processing to accelerate decision-making. However, due to the variability in participant capabilities and task requirements, the complexity of task assignments becomes challenging. This complexity necessitates a mechanism that balances incentives and response time, ensuring tasks are completed within predefined budget and time limits. This paper presents a new task assignment strategy that categorizes participants based on their capabilities and task needs. Using the Hungarian algorithm, our methodology optimizes task assignments with an objective function aiming to minimize both monetary and time costs. We then evaluate the minimum budget needed for successful task completion and its dependency on objective function parameters. A comparison of our method's performance against a standard greedy approach demonstrates its effectiveness. The results suggest that our method enhances the efficiency and reliability of task assignment in EES systems, with potential applications in smart cities, environmental monitoring, and other areas requiring efficient remote sensing.

Edge Computing aims to push services closer to end-users, greatly enhancing latency and scale. Yet, there’s untapped potential beyond the network’s last mile, on the extreme edge. Extreme Edge Computing (XEC) is a computing paradigm that exploits computational resources in the end-user’s immediate vicinity. Edge Crowd Computing (ECC) is an orchestrated sharing economy model within XEC that uses idle resources on user-owned devices for service provision, compensating owners. We analyze an orchestrated ECC where devices rent resources in exchange for incentives. Our Incentive-Vacation Queueing (IVQ) model associates performance with incentive payments using vacation queueing, considering the multi-tenancy of devices through a server vacation dependent on incentives received. In this paper, we offer a framework for analyzing any sharing economy system that can be modeled using IVQ. We discuss the relationship between incentives and vacations on performance, namely the incentive-vacation or IVQ function. We examine two families of IVQ functions that can be adjusted to benefit either the orchestrator or the worker and introduce a performance metric for such preference. We derive analytical expressions for system performance that consider the random nature of worker devices’ availability due to fluctuating incentives. The IVQ model explores commodifying user-owned resources in an ECC system, presenting a general approach for performance analysis in such environments.

In Edge Computing, computation is pushed towards the end-user to reduce backhaul load, address nascent privacy issues, and enable a range of low latency applications. Extreme Edge Service systems (EES) are a subset of Edge Computing in which services are deployed on user-owned devices in the proximity of the end-user. In this work, we model and analyze an orchestrator-based EES in which users' devices are recruited in exchange for an incentive. We propose to model the incentives' impact on performance using Incentive-Vacation Queueing (IVQ), a vacation queueing model in which server vacations are a proxy for incentives. Moreover, we derive closed-form expressions to evaluate the performance and directly link the performance to incentives, showing the impact of each one of the system parameters.

The demand for cloud services is expected to exceed the capacity of the centralized cloud. This rise compelled service providers to decentralize the cloud by physically pushing service provision to the proximity of the end-users, which led to the synthesis of solutions such as Fog and Edge computing. Edge Computing seeks to deploy services in the last mile to the end-user, however there is still opportunity on the edge beyond the last mile: the user's own devices. Extreme Edge Computing (EEC) is an edge sub-paradigm that seeks to tap into the idle computational power on non-enterprise user-owned devices. In this work, we navigate some of the challenges posed by EEC that constrain the usage of resources on user-owned devices. We evaluate an orchestrator-based extreme edge system, that oversees user-owned worker devices, and it provides resources in exchange for an incentive payment. We propose the Incentive-Vacation Queueing (IVQ) model to investigate the performance of user-owned worker devices under a vacation policy that is influenced by incentives. We derive closed-form expressions for the system performance that capture the epistemic uncertainty stemming from unexpected user behavior, to show the impact of each parameter in the system performance, and to optimize it. The IVQ model provides insight into the impact of introducing incentives on the workers' performance.

Edge sensing (ES) is rising as a potential solution for remote sensing challenges, as it exploits the proliferation of smartphones, leverages their embedded sensors to collect data from users' surrounding environments and uses their processors to perform edge computing tasks. Moreover, it is characterized by its low cost and time efficiency. Tremendous efforts have been dedicated to ES systems' quality of data (QoD) and coverage to enhance its performance. Since users incentivization plays a crucial role in enhancing the system's performance, the research community concentrated on improving incentives schemes. In this paper, we evaluate the effect of users' mobility on ES systems' quality of data and coverage, and propose a users' distribution-based dynamic-incentive scheme. In particular, we use a 2-dimensional random waypoint (RWP) model to emulate the randomness of users' mobility and velocity. The proposed incentive scheme aims to eliminate the negative impact of mobility on the QoD; by considering different factors to determine users' incentives and creating users' attraction areas in the targeted cells.

Federated Learning (FL) is a novel distributed learning paradigm in which local learning models are simultaneously trained using the stored data on multiple devices, then ultimately aggregated into a global model. A promising use case of FL is the training of a global model using the data collected by unmanned aerial vehicles (UAVs) during their flight, which is invaluable in scenarios in which an infrastructure cannot be accessed (e.g., disaster). However, this is challenging as limited resources are to be distributed between flight time, sensing, processing, and communication. In this paper, we address the resource problem for a set of heterogeneous UAVs with different computation and communication capabilities from distributed point of view. We propose the usage of Device-to-Device (D2D) communication to fairly distribute the data so-far collected by UAVs with different capabilities by posing it as an optimal transport problem. Our contribution is two-fold: (1) We obtain the fairest distribution of data given the UAVs’ computational capabilities such that global learning time is minimal; (2) We devise a scheme using Optimal Transport (OT) to achieve such a fair distribution between UAVs. The performance of the proposed techniques is demonstrated in an FL setting with different UAV topologies with the FL training done using the MNIST dataset.

Mobile crowdsensing (MCS) is a paradigm that exploits the presence of a crowd of moving human participants to acquire, or generate, data from their environment. As a part of the Internet-of-Things (IoT) paradigm, MCS serves the quest for a more efficient operation of a smart city. Big data techniques employed on this data produce inferences about the participants' environment, the smart city. However, sufficient amounts of data are not always available. Sometimes, the available data are scarce as it is obtained at different times, locations, and from different MCS participants who may not be present. As a consequence, the scale of data acquired may be small and susceptible to errors. In such scenarios, the MCS system requires techniques that acquire reliable inferences from such limited data sets. To that end, we resort to small data (SD) techniques that are relevant for scarce and erroneous scenarios. In this article, we discuss SD and propose schemes to tackle the problems associated with such limited data sets, in the context of the smart city. We propose two novel quality metrics: 1) MAD quality metric (MAD-Q) and 2) MAD bootstrap quality metric (MADBS-Q), to deal with SD, focusing on evaluating the quality of a data set within MCS. We also propose an MCS-specific coverage metric that combines the spatial dimension with MAD-Q and MADBS-Q. We show the performance of all the presented techniques through closed-form mathematical expressions, with which simulation results were found to be consistent.

Following the revolutionary changes the Internet of Things (IoT) has introduced to sensor networks, the Mobile Crowd Sensing (MCS) paradigm aims to utilize people and their smartphones as an extended instrument to sense. However, the benefit of MCS is limited when it comes to microeconomic entities rather than macroeconomic entities. In this paper, we propose CrowdDelegate (CD), an extension of MCS that aims to delegate employee tasks of a consumer hypermarket to customer-workers, utilizing store's loyalty programs interface to recruit participants and reduce operational and logistical costs. This is done by assigning CD tasks to customer-workers present around the store requesting their engagement in a gamified loyalty membership. Customers are rewarded points for the execution of CD activities, allowing the retail business to channel a portion of the loyalty program budget towards the reduction of labor costs. The benefits are two-fold as this approach increases cost-efficiency as well as customer retention. A restricted optimal transport is proposed over the topology of the store to recruit customer-workers based on task costs. This paper sheds light on an unexplored potential of human-centric sensing, extending it to benefit businesses and to engage participants in "doing" instead of only sensing.

Smart cities are becoming more complex and greater volumes of data are required for its efficient operation. Mobile Crowdsensing (MCS) is a paradigm that employs smartphones as instruments to collect data, where the recruitment of participants is based on rewards and incentives. However due to the mobile nature of people, sensing may not be available in a specific area of interest, reducing the quality of the MCS inference of that region. In this paper, we propose a method that utilizes optimal transport so that the MCS administrator could direct participants towards areas with poor quality to improve overall quality. An analysis of optimal transport is presented where the method is evaluated using computer simulations, where it is shown to be efficient for moving participants among spatiotemporal cells.

This paper considers Mobile Crowd-Sensing (MCS) systems that suffer from scarce participant availability due to small sample sizes in each sensing cycle. With such small sample sizes, a sample in error would dramatically affect the MCS system performance. Therefore, we propose a novel quality of source metric targeted for small sample sizes through the non-parametric bootstrap, the trimmed mean, and the Median Absolute Deviation Trimming-based mean (MAD-mean). This statistic permits outlier detection, and therefore allows the estimation of quality under the stringent conditions of small sample sizes present in MCS independent sensing cycles. We introduce an algorithm that allows MCS administrators to control the accuracy of the metric, and therefore control the range of accepted values. Such control is achieved by means of introducing the MAD-mean, which deliberately widens the statistic's distribution, and therefore the perception of quality. In combination with the bootstrap, our metric allows quality estimation for samples as small as 8. We develop our robust quality of source metric algorithm, showing the impact of all the involved parameters; and we compare it to computer simulations to demonstrate its viability.

Air pollution is a major issue contributing to global warming that threaten the quality of life on Earth. Numerous research disciplines are combining their efforts to combat air pollution by developing new methods to monitor and control pollution. For this to happen, researchers need to have instant access to new data. In this paper, we have developed a Semi-Autonomous Unmanned Aerial Vehicle (UAV) loaded with sensors to measure different quantities indicating air pollution, in particular: temperature, humidity, dust, carbon monoxide, carbon dioxide, and ozone. The purpose of this UAV is to automatically patrol high altitudes to obtain sensor readings, and transmit raw data to a centralized server via mobile network for visualization and storage. Actual measurements and data collection is carried out in Qatar. This combination of the UAVs' mobility, remote sensing, and networking facilities allows concerned parties such as researchers, smart city administrators and crowd managers, to view and visualize relevant data with significant ease via a web interface, or an android app.