Federated Learning (FL) enabled creating models that are competitive to centralized Machine Learning models, without compromising user privacy. Participating FL clients train local models on their data and only share model weights. An FL server aggregates these weights into global weights that are pushed to clients for the next training round. Despite FL research growth, most of this work is conceived in experimental simulated environments that don’t reflect its applicability to real-world scenarios. Also, existing open-source FL testbeds/frameworks have drawbacks that prohibit convenient deployment over a large spectrum of heterogeneous clients in realistic environments. These drawbacks include simulations, unrealistic datasets, not supporting heterogeneity, and not having realistic environment control in terms of network and client churn, for example. In this paper, we introduce (RealFL) a novel, realistic, open-source, and extendable platform for FL that supports a large scale of heterogeneous clients. It enables realistic assessment of FL solutions by controlling various environmental parameters; e.g. network, client churn, data distribution, training complexity, and client heterogeneity. Using these parameters, we assess RealFL performance through an extensive evaluation. Preliminary evaluation shows a performance gap of up to 72% in training time and 27% in accuracy between FL simulated environments and RealFL. Moreover, extensive evaluation reveals that realistic environmental parameters could affect accuracy by up to 52.7%, training time by up to 77.5%, and communication overhead by up to 98%.

Federated Learning (FL) enabled creating models that are competitive to centralized Machine Learning models while preserving privacy by allowing clients to train data locally. Despite FL research growth, most of the work assessment and existing open-source FL testbeds/frameworks have drawbacks that prohibit convenient deployment over a large spectrum of heterogeneous clients in realistic environments. These drawbacks include simulations, unrealistic datasets, not supporting heterogeneity, and not having a realistic environment control in terms of network and client churn, for example. In this paper, we introduce (RealFL) a novel, realistic, open-source, and extendable platform for FL that supports a large scale of heterogeneous clients. It enables realistic assessment of FL solutions by controlling various environmental parameters; e.g. network, client churn, data distribution, training complexity, and client heterogeneity. Using these parameters, we assess RealFL performance through an extensive evaluation. The results show a performance gap of up to 77.5% in training time and 23.9% in accuracy between FL unrealistic environments and RealFL.

Unmanned Air Vehicles (UAVs), i.e. drones, have become a key enabler technology of many reconnaissance applications in different fields, such as military, maritime, and transportation. UAVs offer several benefits, such as affordability and flexibility in deployment. However, their limited flight time due to energy consumption is one of the key limitations. Therefore, it is crucial to ensure that UAVs can complete the mission while consuming the least energy possible. In this paper, we propose a novel framework for UAV smart navigation to minimize the time and energy of planning mobile targets visitation. We develop a Deep Reinforcement Learning (DRL) approach to allow the drone to learn the targets’ mobility pattern and build its least energy scanning strategy accordingly. We conduct an initial evaluation of the system and our proposed DRL model policy using simulation. Then, to overcome the time-consuming exploration phase of DRL, we develop a Digital Twin (DT) environment of 3D physics-based simulator, which can be used to train the DRL agent efficiently. We also developed a testbed based on hardware integration with the parrot ANAFI drone to verify the feasibility of the proposed methodology. Our findings confirm that the DRL-based agent can achieve performance close to that of a benchmark policy. Moreover, the testbed experiment validates the practicality of utilizing the DT environment for DRL exploration.

Federated Learning is being hailed as a privacy-preserving machine learning alternative, by allowing models to be distributively trained on source devices owning their data. Most FL solutions, and their assessments, however, assume superior environmental reliability, despite the more realistic variances in environmental factors such as device and network capacity, data distribution, and device churn. As such, we argue in this paper, that there is a growing chasm between current FL assessment setups and the evolving FL assessment needs. Motivated by this chasm, we conduct, to the best of our knowledge, the first empirical measurement study of FL performance given realistic environmental factors. Our study quantifies the impact of these environmental factors on FL performance in terms of training time, accuracy, and communication overhead. Our findings have broad implications for the future development of FL including client admission control and scheduling optimizations.

Federated Learning (FL) was introduced to mitigate centralized Machine Learning (ML) data privacy legislation, which restricts the diversity and performance of ML applications. FL allows training data on its source device, and only model updates can leave the device for learning aggregation at the Cloud/Edge. Despite FL advantages, it is challenging for it to realize a training performance that is comparable to traditional ML, especially in terms of training speed. In this article, we first introduce an original analysis of the systems factors contributing to FL training speed. We then define current FL limitations associated with these factors, propose novel and efficient solution strategies to over-come these limitations, and assess the impact of these strategies through preliminary evaluation. Finally, We identify challenges and future research directions that need to be addressed to realize the benefits of pushing FL limits.

Many novel IoT-based applications demand low latency, large compute resources, and high privacy. These requirements have motivated the emergence of fog and edge computing to complement the low-privacy and high-latency cloud. The intention behind Fog computing is to place computational servers closer to the user, typically within the city’s vicinity, to reduce latency. However, because of the high deployment cost of these servers at scale, and unreliable network infrastructures in many countries or areas, edge computing was proposed. Edge computing advocates leveraging compute resources, typically 0-hops away, on distributed ensembles of colocated devices called FemtoClouds. In this paper, we propose MAESTRO, a system that enables users to offload computational jobs to multiple FemtoClouds in their immediate vicinity. For MAESTRO, we build an integrated architecture that includes two new scheduling algorithms for assigning computing workloads to FemtoClouds. Each of our scheduling algorithms is designed to allow the system to operate more efficiently given poor or strong network infrastructures. We implement a full prototype of our system to assess its performance on our experimental testbed. The results indicate that in communication-challenged environments, our specialized scheduler outperforms state-of-the-art schedulers by up to 55%, while in communication-friendly environments our other specialized scheduler outperforms state-of-the-art schedulers by up to 67%.

The massive amount of data generated by mobile/IoT devices worldwide has been both a motivation and a key enabler for Machine Learning (ML)-based applications. Cloud and Edge platforms have been traditionally leveraged as underlying systems where data for ML applications is processed using “Centralized ML.” Federated Learning (FL) was introduced as an alternative solution that provides comparable accuracy, to centralized-ML, while maintaining user privacy. FL allows training data on its source device and only model updates can leave the device for learning aggregation at the Cloud/Edge. However, current FL solutions have limitations that originate from the training process and current system implementations. Hence, in this paper, we propose leveraging FemtoClouds (i.e. ensembles of co-located mobile/IoT devices) as a complementary system to existing FL approaches. We discuss and assess the benefits that FemtoClouds introduce for both the FL training algorithms, and end devices. These benefits include improvements in training accuracy, convergence speed, training time, and energy efficiency. We finally discuss challenges and future research directions that need to be addressed to realize the benefits of Femto Clouds-based FL.

Federated Learning (FL) was introduced to over-come traditional Machine Learning data privacy concerns, and thus, enable us to gain access to more data. Data owners, clients, are orchestrated by a central FL-server to train data locally and only share their model weights. FL approaches have mainly relied on Cloud and/or Edge to aggregate these model weights and propagate training knowledge across clients. However, several issues hinder the scalability of these approaches, especially in communication-challenged environments. In this paper, we propose a novel semi-distributed system to improve FL training accuracy and time, as well as resource-efficiency at the clients. We leverage co-located clusters of high-end IoT devices, known as FemtoClouds, to propagate training knowledge beyond the Edge. We only leverage Edge/Cloud opportunistically to prop-agate knowledge across FemtoCloud pools. Our evaluation shows that our semi-distributed FemtoClouds system achieves competitive accuracy to state-of-the-art FL approaches, with up to 95% time savings and up to 84% energy savings.

Federated Learning (FL) has enabled Machine Learning (ML) applications to capture a larger spectrum of data by allowing such data to remain on-device, a desirable privacy guarantee in many applications. However, the highly iterative nature of FL optimization algorithms requires low-latency and high-throughput connections to clients. Unfortunately, realistic FL training scenarios include heterogeneous clients that are restricted by computation and communication, thereby slowing down or even failing FL training. In this paper, we propose FedTeams; a trust-based and resource-aware FL system that minimizes training latency, while improving accuracy. To achieve this, we mitigate the risk of straggling and weakly-connected clients by leveraging social trust and allowing these clients to offload their data to more powerful trusted peers that can train on their behalf. In specific, we formulate and solve an optimization problem that leverages the FedTeam’s trust graph and client resource information to optimize the distribution of training and minimize training latency. We evaluate FedTeams in a simulated environment, demonstrating up to a 81.6% decrease in training latency and 11.2% increase in global model accuracy when compared to existing state-of-the-art solutions.

The affordability and deployment-flexibility of Unmanned Air Vehicles (UAVs) have ignited the development of many smart applications, including surveillance, disaster management, and smart farming. Drone's energy consumption is a critical issue and it can be controlled through different factors, depending on the application. One approach is to minimize energy consumption by defining a minimal number of strategic target-coverage locations that the drone needs to traverse and efficiently plan the drone's route through these locations. In this paper, we provide solutions that efficiently allow UAVs to cover multiple targets using their cameras. These solutions identify a minimum set of strategic locations that cover the targets and plan the drone's routes across these locations. We address the problem with the objective of minimizing the total energy consumed by the drone during its mission. We model the problem as mixed-integer programming problem and provide a set of heuristics; with and without target clustering. We evaluate the system using simulations. The results indicate the significance of clustering in minimizing the number of strategic locations and saving the drone's energy. Moreover, flexibility in selecting cluster centers provides further reduction in the strategic locations and energy consumption.

Mobile and IoT devices are becoming increasingly capable computing platforms that are often underutilized. In this paper, we propose RAMOS, a system that leverages the idle compute cycles in a group of heterogeneous mobile and IoT devices that can be clustered to form an edge FemtoCloud. At the heart of this system, we formulate a multi-objective, resource-aware task assignment and scheduling problem. The scheduler runs in two main modes; latency-minimization and energy-efficiency. Under the latency-minimization mode, it strives to maximize the computational throughput of the constructed FemtoCloud while maintaining the energy consumption below an operator specified threshold. Under the energy-efficient mode, it minimizes the total energy consumed in the FemtoCloud while meeting defined tasks deadlines. Due to the NP-Completeness of this scheduling problem, we design a set of heuristics to solve it. We implement a prototype of our system and use it to evaluate its performance and efficiency. Our results demonstrate the system's ability to meet different scheduling objectives while adhering to pre-specified time and energy constraints. Compared to other schedulers, RAMOS achieves 10 to 40 percent completion time improvement under latency minimization mode and up to 30 percent more energy-efficiency under the energy-efficient mode.

Many novel IoT-based applications now require large compute resources, high-privacy, and low-latency. This demand has triggered the rise of fog and edge computing to complement the high-latency and low-privacy cloud. Fog computing provides lower latency by bringing computational servers closer to the user, typically within the city's vicinity. However, due to the high cost of deploying such fog servers at scale, and poor network infrastructures in many countries and areas, edge computing has been introduced. Edge computing argues for leveraging compute resources, typically within a user's immediate environment, on distributed ensembles of devices called FemtoClouds. In this paper, we propose Maestro, a system that aids users by offloading computational jobs from them to multiple FemtoClouds in their immediate vicinity. We propose an integrated architecture for Maestro, which incorporates a new scheduling algorithm that assigns compute tasks to FemtoClouds. We implement a full prototype of Maestro, and evaluate its performance on our experimental testbed, as well as through emulation. Our results show that our system and scheduler outperforms state-of-the-art by up to 55%.

The proliferation of smart IoT/mobile devices has ignited the development of new applications that not only require additional compute resources but add new constraints of privacy and low latency. This trend has motivated the recent edge/fog computing paradigm, which introduces a middle tier of rich resources between cloud and IoT/mobile devices. In this article, we argue for the overlooked comput...

Mobile and IoT devices are becoming increasingly capable computing platforms that are often underutilized. In this paper, we propose a system that leverages the idle compute cycles in a group of heterogeneous mobile and IoT devices that can be clustered to form an edge femtocloud. At the heart of this system, we formulate a task assignment and scheduling problem that strives to maximize the computational throughput of the constructed femtocloud while maintaining the energy consumption below an operator specified threshold. Due to the NP-Completeness of this scheduling problem, we design a set of heuristics to solve this problem. We implement a prototype of our system and use it to evaluate its performance. Our results demonstrate the system's ability to utilize the available compute capacity of a group of mobile and IoT devices while adhering to pre-specified energy constraints. Compared to other schedulers, our scheduler achieves up to 40% performance improvement.

Mobile and IoT devices are becoming increasingly capable computing platforms that are often underutilized. In this paper, we propose a system that leverages the idle compute cycles in a group of heterogeneous mobile and IoT devices that can be clustered to form an edge micro-cloud. At the heart of this system, we formulate a task assignment and scheduling problem that strives to maximize the computational throughput of the constructed micro-cloud while maintaining the energy consumption below an operator specified threshold. Due to the NP-Completeness of this scheduling problem, we design a set of heuristics to solve this problem. We implement a prototype of our system and use it to evaluate its performance and assess its efficiency. Our results demonstrate the system's ability to utilize the available compute capacity of a group of mobile and IoT devices while adhering to pre-specified energy constraints. Compared to other schedulers, our scheduler achieves 10% to 40% improvement in terms of latency minimization, and up to 30% improvement in terms of computational throughput.

While smart devices are increasing in numbers, compute resources, and communication capabilities, current application requirements are often outpacing such hardware improvement. Recently, offloading computation to remote cloud resources or closely located computing resources, such as cloudlets or other mobile and IoT devices have been proposed. In this paper, we propose “Cumulus”, an open source platform for edge cloud computational offloading. Cumulus aims at helping engineers and researchers build, test, and evaluate their edge-computing applications, scheduling methods, etc. We present a generic and flexible architecture of Cumulus as well as a proof-of-concept testbed prototype leveraging a large spectrum of heterogeneous devices, communication methods, and OSs. The testbed consists of standalone worker IoT devices and clustered edge mobile clouds of IoT devices. We enhance Cumulus with a visualization tool that allows real time monitoring of resources used by the devices running on the testbed. We evaluate Cumulus in action by demonstrating its performance under several offloading algorithms in different settings.

Low-cost navigation solutions for indoor environments have a variety of real-world applications ranging from emergency evacuation to mobility aids for people with disabilities. Primary challenges for commercial indoor navigation solutions include robust localization in the absence of GPS, efficient route-planning and replanning techniques, and effective user interfaces for resource-constrained platforms like smartphones and mobile phones. In this chapter, we present an architecture for indoor navigation using an Android smartphone that integrates three core components of localization, map-representation, and user interface towards a robust and effective solution for guiding a variety of users, from sighted to the visually impaired to their intended destination. Specifically, we developed a navigation solution that combines complementary localization algorithms [10] of dead reckoning (DR) and WiFi signal strength fingerprinting (SSI) with enhanced route-planning algorithms to account for the sensory and mobility constraints of the user to efficiently plan safe routes and communicate the route information with sufficient resolution to address the needs of the users. To evaluate the feasibility of our solution, we develop a prototype application on a commercial smartphone and tested it in multiple indoor environments. The results show that the system was able to accurately estimate user location to within 5 m and subsequently provide effective navigation guidance to the user.

An ambulance or fire truck arriving a couple of seconds late can be the difference between life and death for some. As different technologies emerge, various approaches to expediting the movement of emergency vehicles have evolved. Horns, sirens and flashing lights were early attempts that are no longer sufficient in most places to clear traffic on the emergency vehicle's route. In these situations, traffic signal preemption has made it possible to guide traffic to move in favor of clearing the emergency vehicle's route. Early traffic signal preemption approaches depended on direct communication between an emergency vehicle's signal emitter and a corresponding signal receiver on the traffic light it was approaching. Accordingly, the location of the vehicle could be detected. Later, (D)GPS was used to more accurately locate the emergency vehicle. This solution was further enhanced by using efficient or even optimal path planning algorithms to choose the route of the emergency vehicle. In the state-of-the-art in emergency vehicle routing, online static route selection is combined with traffic-light preemption to make emergency vehicle travel faster and safer along the chosen optimal path. In this thesis, we propose an enhancement to the state-of-the- art approaches for reducing the emergency vehicle's travel time. Our hypothesis is that combining traffic signal preemption with dynamic path planning will increase the efficiency of routing an emergency vehicle. We implement a graph version of the D*Lite informed search algorithm to efficiently and dynamically plan optimal paths for the emergency vehicle while taking into consideration the real-time updates of congestion levels and other delays to travel time. To further improve our solution, we propose a traffic light preemption strategy that seeks to ensure fast and safe travel of the emergency vehicle while, as a secondary priority, maximizes other traffic flow through the intersection. We evaluate our hypothesis through analytical experiments using our implementation of D* Lite, and further validate our proposed solution through scenarios developed using the VISSIM specialized microscopic traffic simulator (15). The results validate our hypothesis demonstrating that dynamic path planning can improve travel time under uncertain congestion conditions, and that incorporating an appropriate traffic light preemption mechanism can further improve travel time for an emergency vehicle; potentially saving lives.