Cloud computing has seen widespread adoption because it increases the productivity and efficiency of industries and allows for effective scalability of their business [1]. Guaranteeing performance levels is at the core of cloud services and requires huge computational resources, especially with the latest advances in technologies such as Artificial Intelligence and the Internet of Things [2]. Typically, customers subscribe to agreements where cloud providers ensure specific levels of reliability, availability and responsiveness to systems and applications and describe penalties if the service levels are not met. At the same time, massive computational resources are a cost for providers and have a significant environmental impact, which will increase in the future. It is estimated that the energy consumption of data centres (which host cloud services) will grow from 292 TWh in 2016 to 353 TWh in 2030 [3], and greenhouse gas emissions will increase over 14% in 2040, compared to a 1-1.6% increase in the 2007–2016 [4].

Generating an accurate machine learning (ML) model is of great importance for the Internet of Vehicles (IoV). However, obtaining such a model is challenging due to the fact that subgroups of in-network vehicles receive data from different resources. A worthwhile investment then would be identifying those groups before inferring models. Similarity metrics are widely used to distinguish different groups. However, the efficiency of most existing similarity measurements is at the cost of increased computational complexity and decreased accuracy, making them unsuitable for IoV’s stringent conditions. To address this issue, we propose a computationally efficient method to measure the similarity of different vehicles, where a simplified version of Earth mover’s distance (EMD) is adopted. This distance metric is then embedded into a distributed clustering algorithm to learn the global pattern for vehicular systems. Our algorithm’s overall performance is measured using an asynchronous message delay simulator. Compared to the best algorithm of the state of the art, our proposed algorithm converges slightly slower (by less than 1%) but improves the clustering accuracy by as much as 20% with synthetic data. Additionally, real-world data collected from vehicles validates the efficiency of our proposed algorithm.

It is a challenge to generate an accurate machine learning model in a distributed network due to the increased concern in data privacy and high cost in gathering all raw data. This paper presents an adaptive asynchronous distributed clustering algorithm for agents in wireless network to learn the global models, while the privacy is protected. Moreover, the communication cost and clustering quality can be adaptively balanced. The proposed clustering algorithm does not require the number of clusters to be pre-defined. To improve the accuracy of the global model, we propose a bounding boxes based method to fully utilize the shape information of clusters. In addition, we consider different knowledge levels of agent and different requirements about the global model. In experiments on randomly generated network topologies, we demonstrate that methods which do more extensive clustering in each cycle, and which exchange descriptions of cluster shape and density instead of just centroids and data counts, achieve more consistent clustering, in significantly shorter elapsed time. We also show that the proposed methods can learn the same number of clusters as the ground truth when clusters are well separated from each other.

Providers of cloud computing systems need to manage resources carefully to meet the desired Quality of Service and reduce waste due to overallocation. An accurate prediction of future demand is crucial to allocate resources to service requests without excessive delays. Current state-of-the-art methods such as Long Short-Term Memory-based models make only point forecasts of demand without considering the uncertainty in their predictions. Forecasting a distribution would provide a more comprehensive picture and inform resource scheduler decisions. We investigate Bayesian Neural Networks and deep learning models to predict workload distribution and evaluate them on the time series forecasting of CPU and memory workload of 8 clusters on the Google Cloud data centre. Experiments show that the proposed models provide accurate demand prediction and better estimations of resource usage bounds, reducing overprediction and total predicted resources, while avoiding underprediction. These approaches have good runtime performance making them applicable for practitioners.

In many application problems, including physical security and wildlife conservation, infrastructure must be configured to ensure or deny paths between specified locations. We model the problem as sub-graph design subject to constraints on paths and path lengths, and propose length-bound reachability constraints. Although reachability in graphs has been modelled before in constraint programming, the interaction of positive and negative reachability has not been studied in depth. We prove that deciding whether a set of positive and negative reachability constraints are satisfiable is NP complete. We show the effectiveness of our approach on decision problems, and also on optimisation problems. We compare our approach with existing constraint models, and we demonstrate significant improvements in runtime and solution costs, on a new problem set.

We demonstrate in-cabin WiFi-based sensing in a real vehicle, tracking a passengers breathing rate in real-time.

To find missing people in a remote area, we propose an autonomous unmanned aerial vehicle (UAV) approach which attempts to locate the target by detecting and localising the radio signals produced by a GSM cell phone. By using a low- weight software defined radio and companion computer, the UAV can act as a GSM base station and induce the missing person's device to attempt to make contact. Through the signal strength values and known UAV location, a series of these contact attempts can be used to quickly and accurately localise their position. As the area in which the missing person might be located may be quite large, and the interaction of radio signals with terrain is potentially complex, an efficient search strategy for exploring the area is required in order to reduce time taken to make contact. We make use of a constraint-based graph-based path planning approach to produce a route for the UAV to traverse in the air passing through expected signals from a large number of possible source locations, and demonstrate through experiments the timely identification and localisation of the cell phone.

In this paper we describe a complex optimization application arising in maintenance scheduling, developed in close collaboration with an industrial partner. We have to plan and schedule preventive and corrective maintenance activities at customer sites by a group of traveling repair technicians. A specific property of the problem considered here is a mix of customers in both urban centers and rural areas. This means that travel times between customers must be considered when balancing overall workload for each agent. We discuss a problem decomposition compatible with current management practice, describe different solvers for the individual problem steps, and show results on real-world data from the industrial partner.

In multi-radio multi-channel wireless mesh networks, energy saving mechanisms try to save energy by putting radios into sleep mode. The decision to switch energy states of radios is taken based on parameters like remaining energy or traffic requests at nodes. In IEEE 802.11 power saving mode (PSM), nodes turn off the radios whenever there is no traffic to receive, send or forward. Nodes wake up radios periodically to check if there is any new traffic demand. Due to waking up radios redundantly and a requirement of tight synchronization PSM misses opportunities to save energy in the multi radio scenario. We propose an advanced energy saving method (AESM), where each node makes an independent decision on switching radios states while satisfying QoS requirements for different types of traffic flows. Experimental evaluation shows that AESM reduces energy consumption by 20% over PSM, while also reducing delay and packet loss to maintain QoS for network performance.

Every year over 75 000 firefighters are injured and 159 die in the line of duty. Some of these accidents could be averted if first response team leaders had better information about the situation on the ground. The SAFESENS project is developing a novel monitoring system for first responders designed to provide response team leaders with timely and reliable information about their firefighters' status during operations, based on data from wireless inertial measurement units. In this paper we investigate if Gradient Boosted Trees (GBT) could be used for recognising 17 activities, selected in consultation with first responders, from inertial data. By arranging these into more general groups we generate three additional classification problems which are used for comparing GBT with k-Nearest Neighbours (kNN) and Support Vector Machines (SVM). The results show that GBT outperforms both kNN and SVM for three of these four problems with a mean absolute error of less than 7%, which is distributed more evenly across the target activities than that from either kNN or SVM.

In dense wireless deployments, such as Enterprise WLANs (EWLANs) and home WLANs, interference may occur because of neighbouring WLANs sharing the same unlicensed spectrum. Mechanisms to centrally manage WLAN deployments cannot effectively mitigate the interference caused by hidden terminals (HTs) in WLANs that belong to different organisations. Furthermore, the impact of interference is amplified if it is combined with long-lived TCP traffic flows, which are becoming increasingly commonplace. In this paper, we focus on mitigating the impact of HTs on long-lived TCP flows in home WLANs. In particular, we study the effect of five key factors on long-lived TCP flows under the impact of HTs: packet bursting, backoff mechanisms, maximum number of RTS attempts, capture affect and the number of associated clients with the same Access Point (AP). Extensive simulation results show that a combination between RTS/CTS messages and bursting increases the throughput up to 8× in the presence of HTs. Therefore, we develop a mechanism called joint RTS/CTS with Bursting (RCBurst) that leverages RTS/CTS messages and packet bursting to mitigate the impact of HTs. The simulation results show that RCBurst achieves an improvement of up to 0.3 in Jain's fairness index over the conventional CSMA/CA, without reducing the overall throughput.

Hypergraphical games provides a compact model of a network of self-interested agents, each involved in simultaneous subgames with its neighbors. The overall aim is for the agents in the network to reach a Nash Equilibrium, in which no agent has an incentive to change their response, but without revealing all their private information. Asymmetric Distributed constraint satisfaction (ADisCSP) has been proposed as a solution to this search problem. In this paper, we propose a new model of hypergraphical games as an ADisCSP based on a new global constraint, and a new asynchronous algorithm for solving ADisCSP that is able to find a Nash Equilibrium. We show empirically that we significantly reduce both message passing and computation time, achieving an order of magnitude improvement in messaging and in non-concurrent computation time on dense problems compared to state-of-the art algorithms.

This paper exposes several problems in managing hidden terminals for uplink traffic in rate-controlled environments, and presents solutions to mitigate them. In particular, we focus on scenarios, in which, clients are associated with an access point (AP). The main challenge stems from the negative interactions between rate-control protocols and hidden terminals. To expose the problems, we use a recent channel estimation approach (CEA) to differentiate the reason for packet losses into three categories, noise, congestion and hidden terminals. Our testbed and simulation-based experiments show that the accuracy of hidden terminal estimations using the CEA degrades as MAC-layer ACK frames are sent with relatively high transmission rates. To improve the accuracy of the CEA, the results demonstrate the necessity and cost of making the AP send ACKs based on the minimum ACK rate of all clients. We propose an adaptive scheme that combines both the CEA and RTS/CTS messages. The proposed scheme increases the overall throughput of Minstrel rate-control algorithm by 60% in case of light congested environments. We also proposed a threshold-based adaptive RTS/CTS scheme based on the prior scheme to handle the highly congested environments. The threshold-based adaptive RTS/CTS scheme improves the overall throughput of the adaptive RTS/CTS scheme and Minstrel algorithm between 20-35%. Finally, we propose and evaluate an opportunistic burst scheme, which enforce fairness among clients. Simulation results show that opportunistic bursting outperforms the prior schemes and Minstrel algorithm in Jain's fairness metric (between 0.11 and 0.38) for a realistic given scenario. It also keeps a relatively high overall throughput.

Mobile and wireless networks have long exploited mobility predictions, focused on predicting the future location of given users, to perform more efficient network resource management. In this paper, we present a new approach in which we provide predictions as a probability distribution of the likelihood of moving to a set of future locations. This approach provides wireless services a greater amount of knowledge and enables them to perform more effectively. We present a framework for the evaluation of this new type of predictor, and develop 2 new predictors, HEM and G-Stat. We evaluate our predictors accuracy in predicting future cells for mobile users, using two large geolocation data sets, from MDC [11], [12] and Crawdad [13]. We show that our predictors can successfully predict with as low as an average 2.2% inaccuracy in certain scenarios.

XG-PON requires an effective dynamic bandwidth allocation (DBA) mechanism for upstream traffic to support quality of service for different classes of traffic. We propose X-GIANT, which extends GPON based GigaPON Access Network (GIANT) DBA, with validated optimisations to the originally proposed key parameters - service timers and assured vs non-assured ratio of medium priority traffic. We implement X-GIANT in a standard-compliant XG-PON module designed for the state-of-the-art ns-3 simulator, tune the above key parameters and show that mean-delay and throughput for different classes of traffic obey the XG-PON requirements and respect priorities at both light and heavy upstream loads. We also show that X-GIANT shows better mean-delay performance than Efficient Bandwidth Utilisation (EBU), a recently proposed, GIANT-derived, priority-based DBA mechanism for XG-PON, for all three classes of traffic simulated.

Ride sharing schemes aim to reduce the number of cars in congested cities, while providing the participants with a cheaper alternative to solo driving. To ensure a ride-sharing scheme thrives, it is important to maintain a high participation rate. This requires an adequate balance between drivers and riders. And thus ride matches should be proposed which maximize the number of participants. Different variants of the ride sharing problem have been solved using mixed integer programming. In this paper, we introduce a constraint programming formulation for the problem that uses cumulative constraints with dependencies between trip times. In experiments based on collected trip schedules from four different regions, the constraint model outperforms the MIP model. However, when we change the problem by assuming all drivers have flexible roles, the MIP model allows faster solution times than the CP model.

This paper describes our demonstration of a network repair problem where a robot bridges a gap between two disconnected wireless nodes by searching for a good position and moving there to forward data between the two nodes. It serves to show the potential for our published solutions for automated network repair. A simple Adhoc network consists of two Intel Galileo Gen 2 nodes exchanging messages and an NXT Mindstorm robot with another Galileo on board healing the network connection between the two Galileos in the case the two get disconnected. The demo showcases a solution that employs mobile agents to serve as relays to bridge the connectivity gaps in the wireless network.

Recent studies demonstrate that the performance of a wireless sensor network (WSN) can be improved by deploying multiple sinks in the network. Therefore, in this paper we present different routing protocols for multi-sink WSNs based on the routing protocol for low-power and lossy networks (RPL). Our protocols use different routing metrics and objective functions (OFs). We use the available bandwidth, delay, MAC layer queue occupancy, and expected transmission count (ETX) as the tie-breaking metrics in conjunction with the shortest hop-count metric. Our OFs use the tie-breaking metrics on a greedy or end-to-end basis. Our simulation results demonstrate that the protocols based on the delay, buffer occupancy, and ETX metrics demonstrate best performance, increasing the packet delivery ratio by up to 25% and decreasing the number of retransmissions by up to 65%, compared to a version of the RPL protocol that only uses the hop-count metric. Another key insight is that, using the tie-breaking metrics on a greedy basis demonstrates a slight performance improvement compared to using the metrics on an end-to-end basis. Finally, our results also demonstrate that multiple sinks inside a WSN improve the RPL-based protocol performance.

In many emerging wireless sensor network scenarios, the use of a fixed infrastructure of base stations for data collection is either infeasible, or prohibitive in terms of deployment and maintenance costs. Instead, we consider the use of mobile devices (i.e. smartphones) carried by people in their daily life to collect data from sensor nodes opportunistically. As the movement of these mobile nodes is, by definition, not controlled for the purpose of data collection, synchronization through contact probing becomes a challenging task, particularly for sensor nodes, which need to be aggressively duty-cycled to conserve energy and achieve long lifetimes. This paper formulates this important problem, providing an analytical solution framework and systematically investigating the effective use of contact probing for opportunistic data collection. We present two new solutions, Sensor Node-Initiated Probing (SNIP) and SNIP-Rush Hours, the latter taking advantage of the temporal locality of human mobility. These schemes are evaluated using numerical analysis and COOJA network simulations, and the results are validated on a small sensor testbed and with the real-world human mobility traces from Nokia MDC Dataset. Our experimental results quantify the relative performance of alternative solutions on sensor node energy consumption and the efficacy of contact probing for data collection, allowing us to offer insights on this important emerging problem.

XG-PON is the next-generation standard for passive optical networks operating at 10Gb/s and TCP is the dominant transport protocol of the Internet. In this paper, we present the first performance evaluation of TCP over XG-PON, considering efficiency, fairness, responsiveness, and convergence. The impact of XG-PON's large delay-bandwidth product and asymmetric bandwidth provision are assessed, together with the dynamic bandwidth allocation mechanism. Our state-of-the-art NS3 simulation uses real implementations of three TCP variants (Reno, CUBIC and H-TCP) from the Network Simulation Cradle. Our results highlight several issues that arise for TCP over XG-PON, and emphasise the need for improved awareness of medium access control and scheduling in the context of specific TCP congestion control behaviour.

Reducing the number of cars driving on roads is an important objective for smart sustainable cities, for reducing emissions and improving traffic flow. To assist with this aim, ride-sharing systems match intending drivers with prospective passengers. The matching problem becomes more complex when drivers can pick-up and drop-off several passengers, both drivers and passengers have to travel within a time-window and are willing to switch roles. We present a mixed integer programming model for this switching rider problem, with the objective of minimizing the total distance driven by the population. We exhibit how the potential saving in kilometers increases as the driver flexibility and the density of the distribution of participants increases. Further, we show how breaking symmetries among the switchers improves performance, gaining over an order of magnitude speed up in solving time, and allowing approximately 50% more participants to be handled in the same computation time.

This paper investigates the converged architecture of an LTE mobile network and a PON access network. We identify that the default handover behaviour for LTE is highly inefficient when performed on a backhaul PON tree-topology. We propose a scheme of intelligent opportunistic caching, using the existing resources at the eNBs, to help mitigate this inefficiency. We show that we can achieve a significant reduction in the amount of redundant traffic sent over the PON and that in circumstances when the PON upstream is heavily congested, we can greatly improve the mobile terminal's sustained data rate by up to 50% during the handover process.

Opportunistic data collection in wireless sensor networks uses passing smartphones to collect data from sensor nodes, thus avoiding the cost of multiple static sink nodes. Based on the observed mobility patterns of smartphone users, sensor data should be preforwarded to the nodes that are visited more frequently with the aim of improving network throughput. In this article, we construct a formal network model and an associated theoretical optimization problem to maximize the throughput subject to energy constraints of sensor nodes. Since a centralized controller is not available in opportunistic data collection, data pre-forwarding (DPF) must operate as a distributed mechanism in which each node decides when and where to forward data based on local information. Hence, we develop a simple distributed DPF mechanism with two heuristic algorithms, implement this proposal in Contiki-OS, and evaluate it thoroughly. We demonstrate empirically, in simulations, that our approach is close to the optimal solution obtained by a centralized algorithm. We also demonstrate that this approach performs well in scenarios based on real mobility traces of smartphone users. Finally, we evaluate our proposal on a small laboratory testbed, demonstrating that the distributed DPF mechanism with heuristic algorithms performs as predicted by simulations, and thus that it is a viable technique for opportunistic data collection through smartphones.

Femtocells are low-cost, user deployed base-stations that can alleviate indoor coverage problems. However as femtocell deployment density increases, inter-femtocell interference can severely limit their effectiveness. In this paper, we present a protocol for coordinated auto-configuration of femtocells in a UMTS environment. We use an iterated heuristic search (GRASP) to assign orthogonal CDMA spreading codes in the downlink to avoid interference between overlapping signals. The protocol is compared against a baseline of current practice on a range of simulated problems and shown to consistently provide greatly improved performance for data users, providing up to 480% of the average data rate the baseline provides on certain problems.

The problem of real-time occupancy forecasting for single person offices is critical for energy efficient buildings which use predictive control techniques. Due to the highly uncertain nature of occupancy dynamics, the modeling and prediction of occupancy is a challenging problem. This paper proposes an algorithm for learning and predicting single occupant presence in office buildings, by considering the occupant behaviour as an ensemble of multiple Markov models at different time lags. This model has been tested using real occupancy data collected from PIR sensors installed in three different buildings and compared with state of the art methods, reducing the error rate by on average 5% over the best comparator method.

A Wireless Sensor Network (WSN) is composed of many sensor nodes which transmit their data wirelessly over a multi-hop network to data sinks. Since WSNs are subject to node failures, the network topology should be robust, so that when a failure does occur, data delivery can continue from all surviving nodes. A WSN is k-robust if an alternate length-constrained route to a sink is available for each surviving node after the failure of up to k-1 nodes. Determining whether a network is k-robust is an NP-complete problem. We develop a Constraint Programming (CP) approach for solving this problem which outperforms a Mixed-Integer Programming (MIP) model on larger problems. A network can be made robust by deploying extra relay nodes, and we extend our CP approach to an optimisation problem by using QuickXplain to search for a minimal set of relays, and compare it to a state-of-the-art local search approach.

Wireless Sensor Networks in volatile environments may suffer damage, and connectivity must be restored. The repairing agent must discover surviving nodes and damage to the physical and radio environment as it moves around the sensor field to execute the repair. We compare two approaches, one which re-generates a full plan whenever it discovers new knowledge, and a second which attempts to minimise the required number of new radio nodes. We apply each approach with two different heuristics, one which attempts to minimise the cost of new radio nodes, and one which aims to minimise the travel distance. We conduct extensive simulation-based experiments, varying key parameters, including the level of damage suffered, and comparing directly with the published state-of-the-art. We quantify the relative performance of the different algorithms in achieving their objectives, and also measure the execution times to assess the impact on being able to make autonomous decisions in reasonable time.

Simulation of building evacuations can be a powerful tool for predicting evacuation outcomes, but for this prediction to be useful it must be produced in a timely manner. The building evacuation outcomes are dependent on the movement decisions of the occupants, but simulating all possible combinations of occupant decisions is infeasible. Our contribution is a novel technique using building structure knowledge in the form of a Network Flow Graph to determine where and when occupants might interact with one another. We decompose the problem into non-interacting groups, to be simulated separately, which leads to a significant simulation workload reduction.

The aim of a modern Building Automation System (BAS) is to enhance interactive control strategies for energy efficiency and user comfort. In this context, we develop a novel control algorithm that uses a stochastic building occupancy model to improve mean energy efficiency while minimizing expected discomfort. We compare by simulation our Stochastic Model Predictive Control (SMPC) strategy to the standard heating control method to empirically demonstrate a 4.3% reduction in energy use and 38.3% reduction in expected discomfort.

We develop a set of solution techniques for real-time evacuation guidance of pedestrians during emergency, focusing on evacuation from buildings during a fire. We model the problem as an extension of a dynamic network flow by allowing for nodes and edges to expire over time. This captures evacuation situations where the spreading hazard renders parts of the network unavailable. We formally state the problem, analyze its complexity, develop a set of heuristic approaches and compare their performance against a number of most relevant alternative approaches. We experimentally demonstrate that our heuristics outperform the alternatives and are suitable for real-time use even for large networks.

We introduce ER-MAC, a hybrid MAC protocol for emergency response wireless sensor networks. ER-MAC is designed as a hybrid of the TDMA and CSMA approaches, giving it the flexibility to adapt to traffic and topology changes. It adopts a TDMA approach to schedule collision-free slots. Nodes wake up for their scheduled slots, but otherwise sleep to conserve energy. When an emergency occurs, nodes that participate in the emergency monitoring change their MAC behaviour by allowing contention in TDMA slots. Simulations in ns-2 show that ER-MAC outperforms Z-MAC with higher delivery ratio, lower latency, and lower energy consumption.

This work introduces a dynamic model for the fire emergency evacuation problem. The model extends the concept safety introduced by Barnes et.al. for the situation when the navigation graph is dynamic. The two possible scenarios are described for using the dynamic model with a Wireless Sensor Network for fire emergency evacuation.

Cognitive radio promises better spectrum utilisation through decentralised control, sensing and decision making. The devices are intended to be inexpensive, upgradeable and adaptable. However, introducing upgraded devices into a decentralised environment may cause unexpected results, the upgrades may require a critical mass to show any benefit, and may have an adverse impact on existing devices, which may deter their adoption. We study, in simulation, the gradual introduction of cognitive devices into an existing population using a spectrum access etiquette, in three different scenarios for the device capability. We show that the new devices gain immediate benefits, and that there is always an incentive to upgrade. We show that overall spectrum efficiency is improved. Finally, we show that there is no negative impact on existing devices, and that the introduction of cognitive devices may even improve the success rate of lesser devices.

A facility with frontroom and backroom operations has the option of hiring specialized or cross-trained workers. Cross-trained workers can be switched between the two rooms depending on demand but are more expensive than specialized ones. Assuming stochastic customer arrival and service times, we seek a smallest-cost combination of cross-trained and specialized workers, together with a policy for switching the cross-trained workers between the rooms, which satisfies constraints on the expected customer waiting time and expected number of workers in the back room. A constraint programming approach using logic-based Benders' decomposition is presented. Experimental results demonstrate the strong performance of this approach across a wide variety of problem parameters. This paper provides one of the first links between queueing optimization problems and constraint programming.

With the advent of cognitive radio technology, new paradigms for spectrum access can achieve near-optimal spectrum utilisation by letting each user sense and utilise available spectrum opportunistically while regulating the interference it imposes on other users through interference constraints. However, the simplest and most common forms of such constraints are binary and transmitter-centric, which are often inefficient since they only consider pair-wise sets of transmitters. Hence, we propose a non-binary receiver-centric constraint model for spectrum access in cognitive radio networks. Such a model is in line with the recently proposed interference temperature metric that constraints whole subsets of transmitters, thereby permitting interfering signals to be introduced and enabling additional communication, leading to improved spectrum utilisation. These constraints are easy to generate and check, and are currently being used to devise a co-operative negotiated etiquette for cognitive radios offering heterogeneous services in a wireless office networking scenario.

Effective reuse of Knowledge Bases (KBs) often entails the expensive task of identifying plausible KB-PS (Problem Solver) combinations. We propose a novel technique based on Constraint Satisfaction to enable more rapid identification of incompatible KBs, leaving fewer combinations on which to conduct a thorough investigation. In this paper, we describe our investigation process, its tools, and the latest empirical results applied to non-binary problems that demonstrate our relaxation approach is an effective method for plausibility testing.