In this paper, we propose an efficient framework for monitoring pollution plumes using sensor-equipped drones. Our approach leverages the power of Reinforcement Learning and Mutual Information to strategically plan drone paths in order to maximize the informativeness of the data collected while minimizing communication costs. We propose a multi-agent Independent Q-Learning scheme, where drones act independently but share a global team reward. The reward is calculated based on both the reduction in plume estimation uncertainty and the communication costs. The proposed framework is adaptable to various problem instances, making it suitable for monitoring diverse physical phenomena. We conduct extensive simulations showing the effectiveness of our approach in achieving highquality plume monitoring, with an error in variance estimation ranging from 3% to 5% when compared with ground-truth value. Results also show that our solution offers good compromise between plume estimation and communication costs. This framework outperforms the random-walk approach up to 32.88% and genetic-based heuristics up to 4.2% in terms of total rewards under the proposed scenarios. The proposed framework is advantageous because it excels not only in providing a good solution but also in inferring it in a reasonable time especially compared to a solution provided by genetic-based heuristics.

Providing high-quality predictions is no longer the sole goal for neural networks. As we live in an increasingly interconnected world, these models need to match the constraints of resource-limited devices powering the Internet of Things (IoT) and embedded systems. Moreover, in the era of climate change, reducing the carbon footprint of neural networks is a critical step for green artificial intelligence, which is no longer an aspiration but a major need. Enhancing the energy efficiency of neural networks, in both training and inference phases, became a predominant research topic in the field. Training optimization has grown in interest recently but remains challenging, as it involves changes in the learning procedure that can impact the prediction quality significantly. This article presents a study on the most popular techniques aiming to reduce the energy consumption of neural networks' training. We first propose a classification of the methods before discussing and comparing the different categories. In addition, we outline some energy measurement techniques. We discuss the limitations identified during our study as well as some interesting directions, such as neuromorphic and reservoir computing (RC).

Mobile crowdsensing is a powerful paradigm that takes advantage of low-cost sensors and population density. It allows for large-scale deployments and collection of extensive data, offering a great advantage in multiple fields such as air pollution monitoring, which is a major concern worldwide. Given the mobile nature of the crowd, mobile crowdsensing platforms need to implement adequate route selection/planning solutions to better guide the crowd through the area of interest and maximize the quality of monitoring. In this paper, we propose two route selection algorithms that take into consideration the low accuracy of low-cost sensors in order to find the most informative routes. The similarity-based route selection algorithm aims to maximize spatial coverage by reducing overlaps between participant routes. The cluster-based route selection takes advantage of hierarchical clustering to build groups of similar points of the map according to explanatory variables. We compare the proposed solutions to baseline route selection algorithms, and the results show that our solutions allow for a better estimation while being efficient in terms of travel distance.

Monitoring pollution plumes is a key issue, given the harmful effects they cause. The dynamic of these plumes, which may be important due to meteorological conditions, makes their study difficult. Real-time monitoring in order to obtain an accurate mapping of the pollution dispersion is helpful and valuable to mitigate risks. In this work, we consider a fleet of cooperative drones carrying pollution sensors and operating in order to assess a pollution plume. The latter is assumed to follow a Gaussian process (GP) with varying parameters. For this use case, we propose an efficient approach to characterize spatially and temporarily the plume while optimizing the path planning of drones. In our approach, drones are guided by a deep reinforcement learning (DRL) model called the categorical deep$Q$-network (Categorical DQN) to maximize the plume coverage while considering budget constraints. Specifically, we develop a scalable independent$Q$-learning (IQL) scheme that shares team rewards based on each drone’s deployment relevance and, therefore, ensures cooperation. We evaluate the performance of the plume parameter estimation as well as the maps generated by the GP regression. By testing our framework on several plume scenarios, we show that it offers good results in terms of both estimation quality and runtime efficiency.

In this paper, we address the energy efficiency of neural networks training through data selection techniques. We first study the impact of a random data selection approach that renews the selected examples periodically during training. We find that random selection should be considered as a serious option as it allows high energy gains with small accuracy losses. Unexpectedly, it even outperforms a more elaborate approach in some cases.Our study of the random approach conducted us to observe that low selectivity ratios allow important energy savings, but also cause a significant accuracy decrease. To mitigate the effect of such ratios on the prediction quality, we propose to use a dynamic selectivity ratio with a decreasing schedule, that can be integrated to any selection approach. Our first results show that using such a schedule provides around 60% energy gains on the CIFAR-10 dataset with less than 1% accuracy decrease. It also improves the convergence when compared to a fixed ratio.

The training energy efficiency of deep neural networks became an extensively studied research topic in the last years. Some of the existing approaches seek to reduce the size of the architecture by either starting the training with a large network and pruning it, or by beginning with a seed architecture and then growing it. Instead of compressing the architecture, other approaches aim to reduce the number of training examples through data selection. While various approaches belonging to these two categories have been proposed, only a few works actually conduct energy measurements. Others merely mention potential gains in efficiency or rely on alternative evaluation metrics such as FLOPs. In this paper, we conduct a series of experiments both on a synthetic dataset and on image classification benchmarks in order to compare the impact of pruning, architecture growing and data selection on training energy consumption and prediction quality. Our results show that growing maintains a high prediction quality but brings limited energy gains when the size of the resulting architecture is large. Pruning can offer high gains, but also impacts accuracy, making it more suited for large models. Data selection provides energy gains correlated with the selectivity rate but causes an accuracy loss. We find that the effectiveness of every technique depends on its hyperparameters and on the architecture size.

We consider IoT networks where nodes are able to move to change the network topology and improve area coverage and network performance. We focus on the problem of global optimization where the nodes make use of the predictability of circumstances that affect network operations, such as the communication and sensing ranges, to anticipate future actions that need to be taken so that the correct operation of the network continues to be guaranteed with a minimum global cost. We provide a Mixed Integer Quadratic Program (MIQP)-based solution that minimizes the overall energy consumed over the entire deployment period while maintaining network connectivity and full area coverage. Results show that significant performance enhancement can be obtained when taking predictability into account compared to the case where nodes make decisions based only on their current observations.

Monitoring air pollution plumes in emergency situations (industrial accidents, natural disasters, deliberate terrorist releases, etc.) becomes an issue of utmost importance in our society given the dramatic effects that the released pollutants can cause. Considering these situations, the pollution plume is strongly dynamic leading to a fast dispersion of pollutants in the atmosphere. Thus, the need for real-time response is very strong and a solution to get precise mapping of pollution dispersion is required to mitigate risks. However, monitoring and forecasting air quality in real time in such situations remains a highly challenging endeavour. In this paper, we suggest a systemic approach for monitoring dynamic air pollution based on aerial sensing (sensors mounted on UAVs). The proposed framework consists of a cycle with feedback loop which will constantly combine a spatio-temporal forecasting model based on a convolutional long short term memory (ConvLSTM) network with a data assimilation technique to get accurate pollution maps, while adjusting at each time the trajectories of drones following uncertainty forecasts. Our solution was evaluated and validated using a highly dynamic real world data set namely Fusion Field Trial 2007 (FFT07). The proposed strategy, together with the obtained evaluation results, are presented, and carefully analyzed.

Participatory sensing leverages population density and involves citizens in the collection of extensive data in multiple fields such as air pollution monitoring, enabling large-scale deployments and improving the knowledge of air quality. This study highlights the potential of low-cost sensors through a data analysis of pollutant concentrations collected during multiple sensing campaigns we co-organized using a participatory sensing platform we designed. We first compare the estimation quality of four statistical models and investigate the impact of sampling frequency on the quality of estimation and energy consumption of the nodes using an energy model based on the sensing duty cycle. In addition, we evaluate the capacity of regression models to recover missing data of one sensor based on the other sensors. Results are satisfactory and reveal that a small decrease in the sampling frequency slightly reduces the estimation quality, but in contrast, allows the nodes to operate on a longer period.

In a post-disaster situation, the construction of replacement communication infrastructure is crucial for the success of rescue operations. LTE Device-to-Device Proximity Services and IoT are considered as key enabling technologies for the construction of such replacement networks. Existing techniques rely on smartphones as relay stations to build a replacement broadcast-based network that connects available devices. In many cases, the using of such networks require querying a given type of IoT devices (e.g, surveillance cameras, heart-rate monitors, temperature sensors) depending on network users and service requirements. In such scenarios, incorporating all relays in the broadcast is inefficient and may lead to poor network performance. In this paper, we propose constructing for each service type a sub-network of relay stations that ensure connectivity among IoT devices providing the same service type. The resulting sub-networks ensure an efficient and robust message dissemination, avoiding transmission redundancy, and resulting in higher energy savings as well as high coverage. These properties have been validated by implementing our solution in NS-3 by extending the LTE D2D ProSe module provided by NIST. Obtained results show significant improvements in terms of energy consumption, and packet delivery ratio.

Wireless sensor networks (WSNs) are widely used in environmental applications where the aim is to sense physical phenomena, such as temperature and air pollution. A careful deployment of sensors is necessary in order to get a better knowledge of these physical phenomena while ensuring the minimum deployment cost. In this paper, we focus on using WSN for air pollution mapping and tackle the optimization problem of sensor deployment. Unlike most of the existing deployment approaches that are either generic or assume that sensors have a given detection range, we define an appropriate coverage formulation based on an interpolation formula that is adapted to the characteristics of air pollution sensing. We derive, from this formulation, two deployment models for air pollution mapping using the integer linear programming while ensuring the connectivity of the network and taking into account the sensing error of nodes. We analyze the theoretical complexity of our models and propose the heuristic algorithms based on the linear programming relaxation and binary search. We perform extensive simulations on a dataset of the Lyon city, France, in order to assess the computational complexity of our proposal and evaluate the impact of the deployment requirements on the obtained results.

Air pollution has major negative effects on both human health and environment. Thus, air quality monitoring is a main issue in our days. In this paper, we focus on the use of mobile WSN to generate high spatio-temporal resolution air quality maps. We address the sensors' online redeployment problem and we propose three redeployment models allowing to assess, with high precision, the air pollution concentrations. Unlike most of existing movement assisted deployment strategies based on network generic characteristics such as coverage and connectivity, our approaches take into account air pollution properties and dispersion models to offer an efficient air quality estimation. First, we introduce our proposition of an optimal integer linear program based on air pollution dispersion characteristics to minimize estimation errors. Then, we propose a local iterative integer linear programming model and a heuristic technique that offer a lower execution time with acceptable estimation quality. We evaluate our models in terms of execution time and estimation quality using a real data set of Lyon City in France. Finally, we compare our models' performances to existing generic redeployment strategies. Results show that our algorithms outperform the existing generic solutions while reducing the maximum estimation error up to 3 times.

One of the main concerns of smart cities is to improve public health which is mainly threatened by air pollution due to the massively increasing urbanization. The reduction of air pollution starts first with an efficient monitoring of air quality where the main aim is to generate accurate pollution maps in real time. Spatiotemporally fine-grained air pollution maps can be obtained using physical models which simulate the phenomenon of pollution dispersion. However, these simulations are less accurate than measurements that can be obtained using pollution sensors. Combining simulations and measurements, also known as data assimilation, provides better pollution estimations through the correction of the fine-grained simulations of physical models. The quality of data assimilation mainly depends on the number of measurements and their locations. A careful deployment of nodes is therefore necessary in order to get better pollution maps. In this paper, we tackle the deployment problem of pollution sensors and propose a new mixed integer programming model allowing to minimize the overall deployment cost of the network while achieving a required assimilation quality and ensuring the connectivity of the network. We then design a heuristic algorithm to solve efficiently the problem in polynomial time. We perform extensive simulations on a dataset of the Lyon city, France and show that our approach provides better air quality monitoring when compared to existing deployment methods that are designed without taking into account the outputs of physical models. We also show that in terms of connectivity, the communication range of sensor nodes might have a noteworthy impact on the quality of pollution estimation.

One of the main issues in UAVs networks design is how nodes are relocated in order to meet the desired performance objectives. In this work, we propose two UAVs fleet mobility models based on the Virtual Forces Algorithm (VFA). The application context we are interested in is the air pollution surveillance over wide areas. The first model is a centralized variant where all computations are performed in a central ground base station. While the second model is a distributed version where each node takes its own decision in collaboration with its neighbors. We evaluate our models performances and we compare them with state of the art solutions using a real data set of air pollution concentrations and according to three main metrics: the maximal estimation error, execution time and communication cost.

Wireless sensor networks (WSN) are widely used in environmental applications where the aim is to sense a physical parameter such as temperature, humidity, air pollution, etc. Most existing WSN-based environmental monitoring systems use data interpolation based on sensor measurements in order to construct the spatiotemporal field of physical parameters. However, these fields can be also approximated using physical models which simulate the dynamics of physical phenomena. In this paper, we focus on the use of wireless sensor networks for the aim of correcting the physical model errors rather than interpolating sensor measurements. We tackle the activity scheduling problem and design an optimization model and a heuristic algorithm in order to select the sensor nodes that should be turned off to extend the lifetime of the network. Our approach is based on data assimilation which allows us to use both measurements and the physical model outputs in the estimation of the spatiotemporal field. We evaluate our approach in the context of air pollution monitoring while using a dataset from the Lyon city, France and considering the characteristics of a monitoring system developed in our lab. We analyze the impact of the nodes' characteristics on the network lifetime and derive guidelines on the optimal scheduling of air pollution sensors.

Due to the increasing industrialization and the massive urbanization, air pollution monitoring is being considered as one of the major challenges of smart cities. Many air pollution monitoring systems have been proposed in the literature, among which wireless sensor networks seem to be a leading solution thanks to sensors' low cost and autonomy as well as their finegrained deployment. A careful deployment of sensors is therefore necessary to get better performances while ensuring a minimal financial cost. In this paper, we consider citywide wireless sensor networks and tackle the minimum-cost node positioning issue for air pollution monitoring. We propose an efficient approach that aims to find optimal sensors and sinks locations while ensuring air pollution coverage and network connectivity. Unlike most of the existing methods, which rely on simple and generic detection models, our approach is based on the spatial analysis of pollution data, allowing to take into account the nature of the pollution phenomenon. As proof of concept, we apply our approach on real world data, namely the Paris pollution data, which was recorded in March 2014. We also perform extensive simulations in order to study the performance of our approach in comparison to the existing methods.

Mobile wireless sensor networks (MWSN) are widely used for monitoring physical phenomena such as air pollution where the aim is usually to generate accurate pollution maps in real time. The generation of pollution maps can be performed using either sensor measurements or physical models which simulate the phenomenon of pollution dispersion. The combination of these two information sources, known as data assimilation, makes it possible to better monitor air pollution by correcting the simulations of physical models while relying on sensor measurements. The quality of data assimilation mainly depends on the number of measurements and their locations. A careful deployment of nodes is therefore necessary in order to get better pollution maps. In this ongoing work, we tackle the placement problem of pollution sensors and design a mixed integer programming model allowing to maximize the assimilation quality while ensuring the connectivity of the network. We perform some simulations on a dataset of the Lyon city, France in order to show the effectiveness of our model regarding the quality of pollution coverage.

Air pollution has become a major issue in the modern megalopolis because of industrial emissions and increasing urbanization along with traffic jams and the heating/cooling of buildings. Monitoring urban air quality is therefore required by municipalities and the civil society. Current monitoring systems rely on reference sensing stations that are precise but massive, costly, and, therefore, seldom. In this paper, we focus on an alternative or complementary approach, with a network of low cost and autonomic wireless sensors, aiming at a finer spatiotemporal granularity of sensing. Generic deployment models in the literature are not adapted to the stochastic nature of pollution sensing. Our main contribution is to design integer linear programming models that compute sensor deployments capturing both the coverage of pollution under time-varying weather conditions and the connectivity of the infrastructure. We evaluate our deployment models on a real data set of Greater London. We analyze the performance of the proposed models and show that our joint coverage and connectivity formulation is tight and compact, with a reasonable enough execution time. We also conduct extensive simulations to derive engineering insights for effective deployments of air pollution sensors in an urban environment.

Air pollution has become a major issue of modern megalopolis because of industrial emissions and increasing urbanization along with traffic jams and heating/cooling of buildings. Monitoring urban air quality is therefore required by municipalities and by the civil society. Current monitoring systems rely on reference sensing stations that are precise but massive, costly and therefore seldom. In this ongoing work, we focus on an alternative or complementary approach, using a network of low cost and autonomic wireless sensors, allowing for a finer spatiotemporal granularity of air quality sensing. We tackle the optimization problem of sensor deployment and propose an integer programming model, which allows to find the optimal network topology while ensuring air quality monitoring with a high precision and the minimum financial cost. Most of existing deployment models of wireless sensor networks are generic and assume that sensors have a given detection range. This assumption does not fit pollutant concentrations sensing. Our model takes into account interpolation methods to place sensors in such a way that pollution concentration is estimated with a bounded error at locations where no sensor is deployed.

Monitoring air quality has become a major challenge of modern cities where the majority of population lives. In this paper, we focus on using wireless sensor networks for air pollution mapping. We tackle the optimization problem of sensor deployment and propose two placement models allowing to minimize the deployment cost and ensure an error-bounded air pollution mapping. Our models take into account the sensing drift of sensor nodes and the impact of weather conditions. Unlike most of existing deployment models, which assume that sensors have a given detection range, we base on interpolation methods to place sensors in such a way that pollution concentration is estimated with a bounded error at locations where no sensor is deployed. We evaluate our model on a dataset of the Lyon City and give insights on how to establish a good compromise between the deployment budget and the precision of air quality monitoring. We also compare our model to generic approaches and show that our formulation is at least 3 times better than random and uniform deployment.

The quality of radio communication links decreases with high temperatures. In this paper, we investigate the effect of temperature on percolation-based connectivity in large scale wireless sensor networks and show that more energy can be saved by allowing some nodes to go to deep sleep mode when temperature decreases and links improve. We determine a closed-form formula for a threshold network density equation in function of temperature τ. Beyond λ*(τ), the network percolates and guarantees connectivity. Based on this result, we propose a simple yet efficient Temperature-Aware MAC plugin (TA-MAC) that enables the underlying MAC protocol to dynamically adapt the network effective density to allow further energy savings while maintaining network connectivity. TA-MAC can be potentially used with any wireless MAC protocol. We carried out simulations and demonstrated that BMAC and SCP-MAC augmented with TA-MAC plugin allow a significant energy efficiency improvement.

Recently, air pollution monitoring emerges as a main service of smart cities because of the increasing industrialization and the massive urbanization. Wireless sensor networks (WSN) are a suitable technology for this purpose thanks to their substantial benefits including low cost and autonomy. Minimizing the deployment cost is one of the major challenges in WSN design, therefore sensors positions have to be carefully determined. In this paper, we propose two integer linear programming formulations based on real pollutants dispersion modeling to deal with the minimum cost WSN deployment for air pollution monitoring. We illustrate the concept by applying our models on real world data, namely the Nottingham City street lights. We compare the two models in terms of execution time and show that the second flow based formulation is much better. We finally conduct extensive simulations to study the impact of some parameters and derive some guidelines for efficient WSN deployment for air pollution monitoring.

Temperature variations have a significant effect on low power wireless sensor networks as wireless communication links drastically deteriorate when temperature increases. A reliable deployment should take temperature into account to avoid network connectivity problems resulting from poor wireless links when temperature increases. A good deployment needs also to adapt its operation and save resources when temperature decreases and wireless links improve. Taking into account the probabilistic nature of the wireless communication channel, we develop a mathematical model that provides the most energy efficient deployment in function of temperature without compromising the correct operation of the network by preserving both connectivity and coverage. We use our model to design three temperature-aware algorithms that seek to save energy (i) by putting some nodes in hibernate mode as in the Stop-Operate (SO) algorithm, or (ii) by using transmission power control as in Power-Control (PC), or (iii) by doing both techniques as in Stop-Operate Power-Control (SOPC). All proposed algorithms are fully distributed and solely rely on temperature readings without any information exchange between neighbors, which makes them low overhead and robust. Our results identify the optimal operation of each algorithm and show that a significant amount of energy can be saved by taking temperature into account.

Given the sensitivity of the potential WSN applications and because of resource limitations, key management emerges as a challenging issue for WSNs. One of the main concerns when designing a key management scheme is the network scalability. Indeed, the protocol should support a large number of nodes to enable a large scale deployment of the network. In this paper, we propose a new scalable key management scheme for WSNs which provides a good secure connectivity coverage. For this purpose, we make use of the unital design theory. We show that the basic mapping from unitals to key pre-distribution allows us to achieve high network scalability. Nonetheless, this naive mapping does not guarantee a high key sharing probability. Therefore, we propose an enhanced unital-based key pre-distribution scheme providing high network scalability and good key sharing probability approximately lower bounded by 1-e-1≈ 0.632. We conduct approximate analysis and simulations and compare our solution to those of existing methods for different criteria such as storage overhead, network scalability, network connectivity, average secure path length and network resiliency. Our results show that the proposed approach enhances the network scalability while providing high secure connectivity coverage and overall improved performance. Moreover, for an equal network size, our solution reduces significantly the storage overhead compared to those of existing solutions.

Given the sensitivity of the potential WSN applications, security emerges as a challenging issue in these networks. Because of the resource limitations, symmetric key establishment is one favorite paradigm for securing WSN. One of the main concerns when designing a key management scheme for WSN is the network scalability. Indeed, the protocol should support a large number of nodes to enable a large scale deployment of the network. In this paper, we propose a new highly scalable key establishment scheme for WSN. For that purpose, we make use, for the first time, of the unital design theory. We show that the basic mapping from unitals to pairwise key establishment allows to achieve an extremely high network scalability while degrading the key sharing probability. Then, we propose a new unital-based key pre-distribution approach which provides high network scalability and good key sharing probability. We conduct analytical analysis to compare our solutions to existing ones, the obtained results show that our approach enhances the network scalability while providing good overall performances. Also, we show that our solutions reduce significantly the storage overhead at equal network size compared to existing solutions.

Tree topologies are widely used in WSN in order to route convergecast traffic to the sink. We consider in this paper the Shortest Path routing Tree (SPT) problem in WSN under different metrics; we show that the basic SPT based strategies are unsuitable for the many-to-one WSN when considering some metrics to compute link costs. Indeed, existing SPT approaches aim to construct a tree rooted at the sink such that the cost of the path from any node to the sink is minimal, while the cost of a given path is computed as summation of the costs of links that compose this path. However, in many-to-one WSN, links which are close to the sink are more critical than other links when using some metrics. We propose in this paper a new weighted path cost function, and we show that our cost function is more suitable for WSN. Based on this cost function, we propose a simple and efficient weighted shortest path tree construction which does not introduce new overheads. We consider, then, the particular case of energy-aware routing in WSN when we apply our new solution in order to construct more suitable energy-aware SPT. We conduct extensive simulations which show that our approach allows to enhance the network lifetime up to 17% compared to the basic one.

Given the sensitivity of the potential applications of wireless sensor networks and because of resource limitations, security emerges as a challenging issue in these networks. Key management is one of the required building blocks of many security services. Unfortunately, public key based solutions, which provide efficient key management services, are unsuitable for WSN because of resource limitations. Then, symmetric key establishment is the favorite paradigm for securing exchanges in WSN. In this work, we resume our contributions to enhance the resilience of WSN key pre-distribution schemes through hash chaining. We present our solutions as well as analytical analysis and simulation results which show that our approaches enhance the network resiliency without introducing overheads compared to existing solutions.

We propose, in this paper, a novel class of probabilistic key pre-distribution schemes highly resilient against node capture. We introduce a new approach to enhance resilience by concealing keys through the use of a simple hash chaining mechanism. We provide analytical analysis which shows that our solution enhances the network resilience against node capture without introducing a new overhead comparatively to similar solutions in the literature.

Key management is a corner stone service for any security solution for WSNs. Resources limitation of WSNs makes the public key based solutions, which offer more efficient key management services, unsuitable for wireless sensor networks. In this paper, we propose a novel efficient tree-based probabilistic key management scheme which is highly resilient against node capture attacks. Our solution is based on symmetric cryptography with a probabilistic key pre-distribution. We introduce a new approach to enhance resilience by concealing keys through the use of a simple hash function mechanism. We further improve resiliency by using the number of shared keys as criteria to construct the secure tree. We provide analytical analysis and extensive simulations which show that our solution enhances the network resilience against node capture without introducing any overhead comparatively to similar solutions in the literature.