Understanding the distribution of bird species and populations and learning how birds behave and communicate are of great importance in wildlife biology, animal ecology, conservation of ecosystems, and assessing the effects of climate change and urbanization. The temporal and spatial limitations of human observation have motivated significant efforts to develop technology for bird song and vocalization detection and classification. While solutions based on signal processing and machine learning are extant, they are limited in various combinations of speed, computational complexity, and memory use, as well as in detection/classification capability in real-world conditions. This paper introduces ToucaNet, a deep neural network for birdsong detection based on transfer-learning, a deep learning mechanism allowing us to exploit knowledge acquired on various tasks: this enables us to speed up training and shows improved detection accuracy. ToucaNet provides birdsong detection accuracy in line with the best solutions in the literature but with much less computational complexity and memory demand. We also introduce BarbNet, an approximated version of ToucaNet tailored for Internet-of-Things (IoT) units. We show the proposed solution’s effectiveness and efficiency in terms of detection accuracy and the implementation feasibility in real-world IoT devices, with specific results for the STM32 Nucleo H7 board, which is based on an ARM Cortex-M7 processor. To our best knowledge, this is the first birdsong detection algorithm designed to take into account constraints on memory, computational speed, and power usage of embedded devices. Thus, this work points the way to cost-effective IoT technology for at-scale intelligent birdsong data collection and analysis in the field.

When a disaster strikes a city, conventional information networking infrastructure can be damaged or completely lost, leading to suffering or loss of life. These consequences are avoidable with the introduction of an urban networking fabric for emergency services. We describe here an overall design for a Network for Urban Emergency Services (NUES) that provides essential communication services during disasters and post-disaster recovery. The core of NUES is a dense wireless networking fabric composed of low-cost wireless nodes. Because each node is autonomously powered, and NUES takes advantage of satellite networking services, the overall NUES architecture can be independent of all other (conventional) infrastructure, depending only on users' battery-powered devices (such as cell phones or tablets) and satellite data services. But it also has the ability to exploit conventional infrastructure (in particular cellular wireless) to reduce cost and provide greater bandwidth to users. Since some NUES nodes themselves may be damaged or lost, NUES is also designed for resilience, with redundant nodes connected by a self-healing network. Due to its dense coverage and modular node architecture, NUES also provides smart city sensing services the year around, greatly increasing its cost efficiency.

A district heating grid is a hot water pipeline grid used to transmit waste heat from CHP (Combined Heat and Power) production plants to buildings, where it is used for service water and space heating. Currently, broad use of district heating is limited to Northern Europe and some regions in Central Europe, but the European Union Energy Efficiency Directive 2012/27/EU and Energy Performance of Buildings Directive 2010/31/EU drive the widespread adoption of this technology in member states. Further, these directives push the introduction of significant renewable local energy production into these grids, creating a need for a smart district heating grid that is in many ways analogous to the smart electric grid. One major difference is that in a hot water pipeline grid, there are significant delays in energy propagation, so algorithms developed for the smart electric grid are not directly applicable. In this paper, an auction based heat trade mechanism is proposed between an auctioneer agent representing the CHP plant operator and prosumer agents representing end users with local solar thermal generation capacity. A predictive simulation is used by an auctioneer to account for the said delays in energy production and to determine the market price that best satisfies the performance indicators defined by the CHP plant operator. The proposed multi-agent system is demonstrated by co-simulating it against a model of a part of a Finnish municipality's district heating grid.

The emergence of LED luminaires that can be dimmed rapidly and frequently enables radical innovations in street lighting automation. It is possible to save considerable amounts of energy by adjusting street lighting according to the movements of individual road users. Since the same lights are often used for illuminating both vehicle lanes and sidewalks, it is necessary to detect both vehicles and pedestrians. A simulation of vehicle and pedestrian traffic is interfaced to a distributed street lighting system based on the IEC 61499 standard, in order to quantify the energy saving potential of a street lighting automation system that exploits real-time sensing information of individual road users while conforming to traffic safety related requirements for road lighting.

A smart electric grid incorporating local generation and storage is an area of active research in the industrial automation community. Very little attention has been given to similar possibilities in hot water grids that are used extensively for district heating in several countries, especially in Central and Northern Europe. In this paper, possibilities for heat trade in such a grid are investigated through simulation, considering the thermo-hydraulic properties of such a grid. For the distributed automation of several independent producer-consumers, an event based IEC 61499 solution is presented. The paper proposes a co-simulation environment for a dynamic process simulator and an event based control system, which is applied to the district heating case.

Sensor/actuator networks are becoming hierarchical, with wireless ad hoc networks interconnected via infrastructured wireless (e.g. the cellular network) and the Internet. This trend is being driven by new application domains such as cyber-ecological systems. In these systems, sensor/actuator networks and data centers form a geographically-distributed closed-loop system capable of near real-time sensing and actuation of the physical environment. To manage the complexity of this interconnected sensor/actuator network system, we describe the use of Data Turbine streaming middleware. Data Turbine enables management of real-time streaming data and provides persistent local storage for resilience to unreliable links. Here, we discuss Data Turbine's use in interconnected real-time sensor/actuator networks for closed-loop control scenarios. We describe how sensor/actuator networks and distributed real-time data centers can be interconnected with Java client applications using the Data Turbine API to accomplish closed-loop sensing, inference, control, and actuation in a test-bed cyber-ecological system, as well as discussing challenges and future work.

The Earth's human population and per capita use of resources have grown dramatically in the past century. Rapid conversion of land from wilderness to agricultural and urban use, air and water pollution, drawdown of potable water resources, overfishing, and increased use of fossil-based fuels have changed the Earth's natural systems, as indicated by increasing average global temperatures, invasions of exotic species, modification of the level and chemistry of our oceans, and rapidly declining biodiversity. Examines the impact to the global ecosystems and discusses initiatives to manage and control the terrestrial, aquatic, and martime systems impacted.

Wireless data gathering networks are often tasked to gather correlated data under severe energy constraints. The use of simple channel codes with source-channel decoding can potentially provide good performance with low energy consumption. Here we consider progressive coding in multi-hop networks, where an intermediate node decodes its received noisy codewords. The estimated information is concatenated with the node's own information word and encoded; the resulting progressively-encoded codeword is then transmitted to the next node. In non-progressive coding, the node simply forwards the received noisy codewords along with its own encoded data. Here we compare the performance of two codes with low decoding complexity, Repeat-Accumulate (RA) and Low-Density Parity-Check (LDPC) codes, in combination with two progressive coding schemes. Progressive channel coding uses only channel decoding at the intermediate node, while progressive source-channel coding uses source-channel decoding, exploiting the probabilistic dependency of the information words (caused by the correlation structure of the data) jointly with the deterministic dependency induced by channel coding. Two decoding schemes are considered at the data center: channel decoding only and iterative source-channel decoding. In simulation experiments, we consider a line network topology with systematic RA and LDPC coding. Results show that progressive coding performs better than non-progressive coding, and RA codes perform better with lower computational complexity than LDPC codes, both for channel-decoding-only and iterative source-channel decoding.

Energy-constrained wireless sensor networks often are designed to measure a spatio-temporal process that is correlated in space, time, or both. The goal of these data-gathering networks is a description of the process that provides the required fidelity with a minimum expenditure of energy. Our approach combines (1) channel coding and estimation/decision processing of coded messages for in-network data censoring with (2) estimation of the encoded and censored information at a fusion center where energy is plentiful. Nodes examine their own messages together with messages from preceding nodes, and compute the fidelity of the estimate at the next node as a function of data censoring proposals. Our algorithm exploits redundancy of two types: intra-message redundancy from channel coding, and inter-message redundancy due to spatio-temporal correlation of the samples. This redundancy is used to alleviate overall energy consumption and message congestion near the fusion center by allowing relaying nodes to censor messages that might otherwise be forwarded, if those messages can be inferred from other messages, given the correlation model. The effect of censoring on fidelity and energy consumption is characterized, and our censoring algorithm shown to provide significant energy savings.

This paper explores integrated source-channel decoding, driven by wireless sensor network applications where correlated information acquired by the network is gathered at a destination node. The collection of coded measurements sent to the destination, called a source-channel product codeword, has redundancy due to both correlation of the measurements and the channel code used for each measurement. At the destination, source-channel (SC) decoding of this code combines decoding using (i) the deterministic structure of the channel-coded individual measurements and (ii) the probabilistic structure of a prior model, called the global model, that describes the correlation structure of the SC product codewords. We demonstrate the utility of SC decoding via MAP SC decoding experiments using a (7,4,3) Hamming code and a Gaussian global model. We also show that SC decoding can exploit even the simplest possible code, a single-parity check code, using a MAP SC decoder that integrates the parity check constraint and global model. We describe the design of a low-complexity message-passing decoder and show it can improve performance in the poor-quality channels often found in multi-hop wireless data-gathering sensor networks.

Even for a specific application, the design space of wireless sensor networks is enormous, and traditional disciplinary boundaries are disappearing in the search for efficient integrated network architectures and protocols. There is a strong need to develop objective frameworks for the evaluation of performance and energetic cost as a function of network control at multiple levels, including the signal/data processing application, network organization, routing, MAC, and physical layers. This paper, a step in this direction, addresses the efficiency of linear estimation of a second-order random spatial field at a central server - a snapshot - in terms of the precision of the estimate and the energetic cost of computing it. We present a model that is based on the tasks of taking samples at a specific resolution and reporting them over the network to the server where the estimate is computed. It provides insight into the joint design of sampling and routing, an explicit efficiency measure for finite networks, and a performance benchmark for precision. The model, with predefined sampling locations at each resolution, implicitly requires centralized control to obtain spatially uniform sampling. We describe a decentralized reporting protocol called PROSE (Protocol for Randomized Opportunistic Sampling and Estimation) that uses localized decisions for sampling, data aggregation, and routing that are random and opportunistic. Finally, we show that the performance of PROSE compares favorably with the benchmark, and that PROSE admits simpler nonopportunistic random behavior at the cost of lower performance.

Bacteria are highly efficient agents that sense, compute, and actuate. Moreover, they can form robust interspecies networks - bacterial communities - via sophisticated communication protocols. We assert that the improved understanding of these communities in the last decade provides a new model for swarm intelligence with distinct advantages, including ease of laboratory experimentation, explicit coupling of communication and behavior, and intergenerational dynamics. The first part of this paper provides a brief overview of bacterial communities in the context of swarm intelligence. The second part describes a promising new application of SI principles inspired by bacterial communities: the design of robust networked embedded and real-time systems.

A fundamental tradeoff in MAC protocol design for wireless ad hoc networks is between proactive and reactive coordination, where the latter is used to resolve access conflicts whose severity is determined by the former. In this paper, we describe clique-based randomized multiple access (CRMA), a distributed MAC protocol for wireless ad hoc network applications. Of the many objectives in MAC design for this application, CRMA places strongest emphasis on energy efficiency and reliance only on local (one-hop) connectivity information. CRMA forms collection of nodes, or cliques, separated by one hop, and provides the proactive coordination required for clique members to synchronize their wake-sleep cycles. Each clique selects a slot in the clique's frame pseudo-randomly, so that no proactive coordination between cliques required. To limit potential access conflicts CRMA can exploit bandwidth via frequency hopping or spread spectrum coding; these also provide robustness to multi-channel radios to increase performance. With a slight amount of additional proactive coordination, CRMA can also employ what we call predictive conflict resolution, wherein clique members predict access conflicts and resolve them ahead of time.

The maximum-likelihood (ML) detector for direct-sequence spread-spectrum signaling over ISI channels is developed for both symbol-length and long spreading sequences. The approach is based on a delay-domain orthogonal decomposition of a joint specular/diffuse dispersive multipath channel. The detector, which extends the Ungerboeck symbol-spaced structure, demonstrates the path diversity inherent in spread spectrum ML detection. The RAKE detector is shown to be the ML detector under a set of ideal conditions, one of which cannot be obtained in practice. As an intermediate result, the DS/SS decision-feedback equalizing detector is shown to be a suboptimum approximation of the ML detector.

In this paper, we develop a physical layer-based framework for the organization of an ad-hoc wireless network. The focus is quasi-static environments, such as multimedia classrooms, and situations characterized by real-time services and high traffic loads. In these cases, centralized control may be preferable due to its simplicity and high efficiency. Using a link loss matrix, an approach is proposed for selection of a network leader that takes into consideration link losses and transmitter powers. Both minimax and minisum criteria are developed to determine uplink, downlink, and overall leaders. A QoS-based iterative algorithm for determination of link transmit powers is also proposed, from which leader selection is obtained as a special case. Numerical results are also presented which explore the effect of quantization of feedback information on the iterative algorithm.

The authors present a description and performance analysis of a new PN code acquisition approach for the tracking and data relay satellite system with multiple access (TDRSS MA) return subsystem. The MA subsystem uses a phased array on-board the TDRS with 30 elements whose received signals are each independently retransmitted to the ground for beam forming. In the current operational system, acquisition is performed by using a serial search of the possible PN epochs on the beam-formed 30-element signal. In contrast, the new approach performs acquisition on a single (arbitrary) element signal using a parallel (i.e., simultaneous) search of all possible PN epochs. Since only one element signal is used, no beam forming is required prior to PN acquisition. The approach is based on sampling and digital processing of the received signal. The new method also incorporates techniques which counteract the effects of Doppler and asynchronous data modulation. The probability of correct acquisition as a function of acquisition time, signal-to-noise ratio, data rate, and user satellite Doppler is analyzed.>

The performance of a digital delta modulator when used as a repeater for binary non-return-to-zero signals between a noisy channel and an error-free link is analyzed. A Markov chain is used to model the accumulator output (observation process) of the device, and a Bayes test with a probability of error criterion is formulated to evaluate its performance in additive band-limited Gaussian channel noise. The key design parameters of sampling rate and accumulator range are varied for computation of performance degradation relative to the optional matched filter. Numerical results are presented that demonstrate the feasibility of this technique.>