Lack of proper real-world data is a fundamental issue of deployed edge AI systems, which can lead to inaccurate AI model performance and, eventually, to system failures. Incremental learning (IL), in which the model is retrained on newly incoming data, can be used to overcome this issue. This study proposes a novel comprehensive self-supervised IL framework for resource-constrained image analysis applications. It focuses on insect segmentation and counting, fundamental operations for edge Internet of Things (IoT) devices deployed in pest control applications. The framework distributes actions between edge nodes, relying on low-power microcontrollers (MCU), and an edge server, implemented on a more powerful processor device, such as a Raspberry Pi. Segmentation and counting tasks are performed on edge nodes using a lightweight deep-learning (DL) model, while the edge server autonomously handles image reconstruction, mask generation, and pseudo labeling to incrementally retrain the AI model. This architectural concept eliminates the need for continuous manual labeling and enables continuous retraining on the edge without transferring large amounts of data to the cloud. Through a series of experiments, we demonstrate accuracy comparable to the manual annotation of images by humans, along with approximately a 10% improvement in segmentation and a threefold reduction in counting error over the base model. Additionally, the results reveal low retraining time (363.3 s) and power consumption (1.2 A at 5.1 V) on an embedded edge server, demonstrating that the framework is extremely suitable for agricultural setups lacking a fixed power supply or network connection, especially given that retraining is only needed occasionally.

Weed classification is a fundamental perception task for agricultural robots and an essential enabler of precision and sustainable farming. Existing solutions often rely on high-power edge computing platforms, which limit long-term autonomous operation in Internet of Things (IoT) environments. Meanwhile, the computational complexity of high-accuracy deep learning models hinders their deployment on resource-constrained micro-controllers (MCUs), a critical component of IoT edge nodes. To address these challenges, this paper proposes a TinyML-enabled energy-efficient IoT framework for on-device weed classification, integrating a novel Three-Dimensional Alignment Knowledge Distillation (TDA-KD) strategy with a lightweight multi-layer dilated-convolution student network. The framework enhances knowledge transfer by jointly aligning (i) individual predictions, (ii) inter-sample correlations, and (iii) class semantics, further strengthened through a multi-temperature calibration mechanism. Experimental results on the DeepWeeds and 4Weeds datasets demonstrate that the proposed student model achieves over 95% classification accuracy with only 240K parameters and 87.51 MFLOPs. The model is successfully deployed on an OpenMV H7 Plus board with an STM32H7 MCU, requiring just 105.68 KB Flash memory and achieving an inference time of 378.3 ms with 510.7 mJ energy consumption per sample. A runtime analysis on the Vitirover horticultural robot shows that, compared with a Jetson Nano-based implementation, the proposed IoT pipeline extends operational time by approximately 30.5%. These results highlight the feasibility of deploying high-accuracy weed classification directly on ultra-low-power IoT devices, thereby significantly enhancing the autonomy, energy efficiency, and scalability of agricultural robots.

The concept of optical transparency in antennas for epidermal electronics is demonstrated in this work as a means of improving the long-term comfort-of-wear level and possibly opening up a wider range of applications. In contrast to previous attempts, the epidermal antenna transparency is achieved by employing dielectric and conductive materials that are both transparent and flexible (i.e., polydimethylsiloxane-transparent conductive textile composite) via a non-clean room procedure that is relatively simpler and less expensive. To demonstrate the concept, a modified rectangular loop epidermal antenna for an arm-worn wireless sensing system operating at 868 MHz Ultra High Frequency (UHF) band is designed. Through a systematic numerical investigation, an interesting radiation response of the loop epidermal antenna as the result of two opposing mechanisms of radiation and loss is revealed, which dictates a specific design guideline for the loop when attached to the body compared to that in free space. Two antenna prototypes were fabricated with the developed transparent composite and its non-transparent counterpart. Then, comprehensive characterizations comparing both epidermal antenna prototypes were carried out, including antenna return loss and far-field tests on a human forearm phantom, and indoor wireless connectivity tests using a human test subject. By showing similar performance between the two prototypes, the study provides a convincing demonstration of the applicability of the developed transparent composite for the class of epidermal antenna and the capability of a transparent antenna to enable wireless connectivity in the context of epidermal electronics.

Effective insect pest monitoring is a vital component of Integrated Pest Management (IPM) strategies. It helps to support crop productivity while minimising the need for plant protection products. In recent years, many researchers have considered the integration of intelligence into such systems in the context of the Smart Agriculture research agenda. This paper describes the development of a smart pest monitoring system, developed in accordance with specific requirements associated with the agricultural sector. The proposed system is a low-cost smart insect trap, for use in orchards, that detects specific insect species that are detrimental to fruit quality. The system helps to identify the invasive insect, Brown Marmorated Stink Bug (BMSB) or Halyomorpha halys (HH) using a Microcontroller Unit-based edge device comprising of an Internet of Things enabled, resource-constrained image acquisition and processing system. It is used to execute our proposed lightweight image analysis algorithm and Convolutional Neural Network (CNN) model for insect detection and classification, respectively. The prototype device is currently deployed in an orchard in Italy. The preliminary experimental results show over 70 percent of accuracy in BMSB classification on our custom-built dataset, demonstrating the proposed system feasibility and effectiveness in monitoring this invasive insect species.

The explosive growth of wireless sensor platforms and their emerging wide range of application areas make the development of a sustainable and robust power source, an essential requirement to enable widespread deployment of these wireless devices. As a solution to this cardinal issue, this article reports the design and fabrication of a resonant vibration energy harvester (VEH) that comprises interleaved springs, manifesting a concertina-shaped structure that can enable large mechanical amplitudes of oscillation. Within a relatively small footprint (9 cm3), this concertina-VEH yields a large power density of 455.6$\mu \text{W}$/cm3g2 while operating at a resonant frequency of 75 Hz. Additionally, the feasibility of the implemented VEH to support near field communication (NFC)-based wireless sensor platform, that is yet uncharted, is also investigated in this work. A very low-power consumption NFC wireless sensor node has been designed and developed for this purpose. The developed concertina VEH has been employed to power the electronics interface of this NFC sensor. Using mechanical energy derived from as low as 0.2-g excitation, our study shows that the VEH can enhance the electromagnetic interaction between the transmitting antenna and the reader, resulting in a 120% increase in wireless communication range for the NFC sensor node. Such a high-performance energy harvester-assisted NFC sensor node has the potential to be used in a wide range of Internet of Things (IoT) platforms as a reliable and sustainable power solution.

This paper presents a long range wide area network (LoRaWAN) testbed for environmental monitoring of artefacts within a museum storage facility. The goal is to identify the optimum feasible wireless technology for this application by studying eight different wireless technologies. A testbed network was deployed inside a 5600 m2concrete building to validate the performance of the candidate wireless technologies by way of measurements. In addition, a LoRaWAN scalability approach was also used to simulate the packet delivery ratio for a 500 node network. The wireless communication performance of LoRa WAN was shown to offer the most optimal solution for wireless communication for museum artefact monitoring application.

An important requirement for Internet of Things applications is the ability to provide fast and energy efficient data collection from wireless sensors. When sensor nodes are located far from the data collection point, currently available long range protocols present challenges associated with a very low data rate and often unreliable connections resulting in excessive energy consumption related to data transmission. To address this problem, we propose a simple and energy-efficient data collection architecture for smart agricultural purposes which require wireless sensing. The architecture involves data collection from nodes located in remote fields or on animals leveraging off the use of drones as a data collection mechanism. In particular, drones can fly over the desired areas (points) and collect high volumes of data that would be otherwise difficult to transfer directly to the sink in a reasonable amount of time and using reasonable amounts of energy. We describe the different components and stages that constitute the proposed architecture emphasizing the networking component. We propose the use of different communication technologies, such as LoRa and WiFi, depending on the data collection requirements. We present an in-lab development of this architecture as a proof-of-concept as well as preliminary results for the architecture. The results reveal that the proposed solution is potentially capable of achieving data collection at high volume, however, the performance does not consider the highest spreading factors of LoRa.

Transport systems incorporating linear synchronous motors (LSMs) enable linear motion at high speed for emerging factory automation applications. The goal of this work is to determine the feasibility of harvesting energy directly from an operational LSM transport system employed in high volume manufacturing. Microelectromechanical (MEMs) based sensor technology, deployed as part of a wireless cyber physical system (CPS), perform near real-time magnetic field measurement for a mobile LSM vehicle. The vehicle under study is purposed for mobile factory automation and is not wired for communications nor does it have an onboard power source. A series of experiments were designed and conducted to establish the magnetic profile of the system. Empirical data capture was conducted on a cycled LSM test-bed comprising of 2 shuttles and 2 x 3 meter lengths of LSM track (MagneMotion QuickStick®100). Varying vehicle speeds were incorporated in the experimental regime to determine how changes in velocity would impact the magnetic profile of the vehicle. The recorded magnetic field data was analysed and a relationship between LSM vehicle speed and magnetic field frequency was established. The study highlights the potential to employ a single receiving coil to enable energy recovery which in turn could power a cyber-physical system (CPS) tasked with performing condition based monitoring of the LSM transport vehicles. This in turn can form the basis for the development of a predictive maintenance system, deployed to an LSM based transport layer in high volume manufacturing environments.

LoRaWAN is a recently proposed MAC layer protocol which manages communications between LoRa-based gate-ways and end-devices. It has attracted much scientific attention due its physical layer characteristics, but mainly due to its versatile configuration parameters. However, it is known that LoRaWAN-based transmissions suffer from extensive collisions due to the unregulated access to the medium. For this reason, various techniques that alleviate the burst of collisions have been proposed in the literature. In this paper, we deal with the problem of fast data delivery in LoRa-based networks. We model a network where transmissions follow a Poisson process. We compute the average packet success probability per Spreading Factor (SF) assuming orthogonal transmissions. We, then, formulate an SF optimization problem to maximize the success probability given an amount of data per node and a maximum data collection time window. We show - both theoretically and using simulations - that the overall success probability can be improved by approximately 100% using optimal SF assignments. We validate our findings using a 10-node testbed and extensive experiments. Despite that experiments reveal the existence of inter-SF interference, our solution still provides the best performance compared to other LoRaWAN configurations.

LoRa-based transmissions suffer from extensive collisions even for low node numbers due to unregulated access to the medium. In order to tackle this problem, we propose a collision-free time-slotted scheduling approach where each node autonomously decides when to transmit a packet based on its unique identifier which is converted to a slot number using a modulo operation. We report through simulations and real experiments that this approach can provide very high reliability when the nodes are synchronized. Moreover, it does not require any additional communication overhead apart from a broadcast packet emitted by the gateway. Our comparison with the native LoRa, as well as to a slotted-LoRa version, shows significant performance gains in terms of packet delivery ratio, especially in the case of low node populations.

Hand motion tracking traditionally requires highly complex and expensive systems in terms of energy and computational demands. A low-power, low-cost system could lead to a revolution in this field as it would not require complex hardware while representing an infrastructure-less ultra-miniature (∼ 100μm — [1]) solution. The present paper exploits the Multiple Point Tracking algorithm developed at the Tyndall National Institute as the basic algorithm to perform a series of gesture recognition tasks. The hardware relies upon the combination of a stereoscopic vision of two novel Lensless Smart Sensors (LSS) combined with IR filters and five hand-held LEDs to track. Tracking common gestures generates a six-gestures dataset, which is then employed to train three Machine Learning models: k-Nearest Neighbors, Support Vector Machine and Random Forest. An offline analysis highlights how different LEDs' positions on the hand affect the classification accuracy. The comparison shows how the Random Forest outperforms the other two models with a classification accuracy of 90–91 %.

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 this paper, we present the design and development of a smart insole that may be used to assess long term chronic conditions that affect the elderly population such as Stroke, Dementia, Parkinson's disease, Cancer, Cardiac Disease and Diabetes. This smart insole offers the potential for evidence base rehabilitation. The ICT solution detect the plantar foot pressure in a free living context through the integration of piezo sensors, microcontroller and Bluetooth technology to empirically measure the pressure at important pressure points. The insole consists of 32 piezo sensors, 01 tri-axial accelerometers, temperature sensor and force sensor to automatically switch ON/OFF the insole. The accelerometers provide context for orientation. The design comprises two flexible PCBs encased in a padded layer, in order to protect the sensors and provide comfort to wearer.

Rheumatoid Arthritis (RA) is a disease which attacks the synovial tissue lubricating skeletal joints. This systemic condition affects the musculoskeletal system, including bones, joints, muscles and tendons that contribute to loss of function and Range of Motion (ROM). Traditional measurement of arthritis requires labour intensive personal examination by medical staff which through their objective measures may hinder the enactment and analysis of arthritis rehabilitation. This paper presents the development of a smart glove to facilitate this rehabilitative process through the integration of sensors, processors and wireless technology to empirically measure ROM. The Tyndall/University of Ulster glove uses a combination of 20 bend sensors, 16 tri-axial accelerometers and 11 force sensors to detect joint movement. All sensors are placed on a flexible PCB to provide high levels of flexibility and sensor stability. The system operation means that the glove does not require calibration for each glove wearer.

Rheumatoid Arthritis (RA) is a disease which attacks the synovial tissue lubricating skeletal joints. This systemic condition affects the musculoskeletal system, including bones, joints, muscles and tendons that contribute to loss of function and Range of Motion (ROM). Traditional measurement of arthritis requires labour intensive personal examination by medical staff which through their objective measures may hinder the enactment and analysis of arthritis rehabilitation. This paper presents the development of a smart glove to facilitate this rehabilitative process through the integration of sensors, processors and wireless technology to empirically measure ROM. The Tyndall/University of Ulster glove uses a combination of 20 bend sensors, 16 tri-axial accelerometers and 11 force sensors to detect joint movement. All sensors are placed on a flexible PCB to provide high levels of flexibility and sensor stability. The system operation means that the glove does not require calibration for each glove wearer.

This paper presents abstraction techniques and modeling approaches to include physical-level antenna and receiver performance effects in a system-level network simulation for Wireless Body Area Networks (WBAN). The simulation platform is based on SystemC which can be used to model digital HW and SW aspects of an embedded application. By using the SystemC Network Simulation Library also a distributed network scenario can be simulated. Here, this platform has been extended to take into account the bit error rate and the path loss associated with antenna positioning in close proximity to the human body and with the design parameters of the wireless receiver. Antenna effects are modeled through a database of performance values based on physical measurements on a human phantom. Path loss information is fed in the SystemC simulator to model the received signal strength as a function of the position of nodes. The same information is used in the physical-level simulation of the receiver to extract bit error rate curves to be used by the SystemC simulator to build a statistical model of packet corruption. Instead of physical-level details, their effects are modeled in a parametric way into the system-level simulation thus combining speed and fidelity and allowing cross-domain design space exploration.

Body Sensor Network (BSN) technology is seeing a rapid emergence in application areas such as health, fitness and sports monitoring. Current BSN wireless sensors typically operate on a single frequency band (e.g. utilizing the IEEE 802.15.4 standard that operates at 2.45GHz) employing a single radio transceiver for wireless communications. This allows a simple wireless architecture to be realized with low cost and power consumption. However, network congestion/failure can create potential issues in terms of reliability of data transfer, quality-of-service (QOS) and data throughput for the sensor. These issues can be especially critical in healthcare monitoring applications where data availability and integrity is crucial. The addition of more than one radio has the potential to address some of the above issues. For example, multi-radio implementations can allow access to more than one network, providing increased coverage and data processing as well as improved interoperability between networks. A small number of multi-radio wireless sensor solutions exist at present but require the use of more than one radio transceiver devices to achieve multi-band operation. This paper presents the design of a novel prototype multi-radio hardware platform that uses a single radio transceiver. The proposed design allows multi-band operation in the 433/868MHz ISM bands and this, together with its low complexity and small form factor, make it suitable for a wide range of BSN applications.

This paper presents the design, implementation and characterization of an energy-efficient smart power unit for a wireless sensor network with a versatile nano-Watt wake up radio receiver. A novel Smart Power Unit has been developed featuring multi-source energy harvesting, multi-storage adaptive recharging, electrochemical fuel cell integration, radio wake-up capability and embedded intelligence. An ultra low power on board microcontroller performs maximum power point tracking (MPPT) and optimized charging of supercapacitor or Li-Ion battery at the maximum efficiency. The power unit can communicate with the supplied node via serial interface (I2C or SPI) to provide status of resources or dynamically adapt its operational parameters. The architecture is very flexible: it can host different types of harvesters (solar, wind, vibration, etc.). Also, it can be configured and controlled by using the wake-up radio to enable the design of very efficient power management techniques on the power unit or on the supplied node. Experimental results on the developed prototype demonstrate ultra-low power consumption of the power unit using the wake-up radio. In addition, the power transfer efficiency of the multi-harvester and fuel cell matches the state-of-the-art for Wireless Sensor Networks.

This work performs an extensive charterisation of precision targeted throwing in professional and recreational darts. The goal is to identify the contributing factors for lateral drift or throwing inaccuracy in the horizontal plane. A multi technology approach is adopted whereby a custom built body area network of wireless inertial measurement devices monitor tilt, force and timing, an optical 3D motion capture system provides a complete kinematic model of the subject, electromyography sensors monitor muscle activation patterns and a force plate and pressure mat capture tactile pressure and force measurements. The study introduces the concept of constant throwing rhythm and highlights how landing errors in the horizontal plane can be attributable to a number of variations in arm force and speed, centre of gravity and the movements of some of the bodies non throw related extremities.

Colour is everywhere in our daily lives and impacts things like our mood, yet we rarely take notice of it. One method of capturing and analysing the predominant colours that we encounter is through visual lifelogging devices such as the SenseCam. However an issue related to these devices is the privacy concerns of capturing image level detail. Therefore in this work we demonstrate a hardware prototype wearable camera that captures only one pixel - of the dominant colour prevelant in front of the user, thus circumnavigating the privacy concerns raised in relation to lifelogging. To simulate whether the capture of dominant colour would be sufficient we report on a simulation carried out on 1.2 million SenseCam images captured by a group of 20 individuals.

For most wireless sensor networks, one common and major bottleneck is the limited battery lifetime. The frequent maintenance efforts associated with battery replacement significantly increase the system operational and logistics cost. Unnoticed power failures on nodes will degrade the system reliability and may lead to system failure. In building management applications, to solve this problem, small energy sources such as indoor light energy are promising to provide long-term power to these distributed wireless sensor nodes. This article provides comprehensive design considerations for an indoor light energy harvesting system for building management applications. Photovoltaic cells characteristics, energy storage units, power management circuit design, and power consumption pattern of the target mote are presented. Maximum power point tracking circuits are proposed which significantly increase the power obtained from the solar cells. The novel fast charge circuit reduces the charging time. A prototype was then successfully built and tested in various indoor light conditions to discover the practical issues of the design. The evaluation results show that the proposed prototype increases the power harvested from the PV cells by 30% and also accelerates the charging rate by 34% in a typical indoor lighting condition. By entirely eliminating the rechargeable battery as energy storage, the proposed system would expect an operational lifetime 10--20 years instead of the current less than 6 months battery lifetime.

At a time when technological advances are providing new sensor capabilities, novel network capabilities, long-range communications technologies and data interpreting and delivery formats via the World Wide Web, we never before had such opportunities to sense and analyse the environment around us. However, the challenges exist. While measurement and detection of environmental pollutants can be successful under laboratory-controlled conditions, continuous in-situ monitoring remains one of the most challenging aspects of environmental sensing. This paper describes the development and test of a multi-sensor heterogeneous real-time water monitoring system. A multi-sensor system was deployed in the River Lee, County Cork, Ireland to monitor water quality parameters such as pH, temperature, conductivity, turbidity and dissolved oxygen. The R. Lee comprises of a tidal water system that provides an interesting test site to monitor. The multi-sensor system set-up is described and results of the sensor deployment and the various challenges are discussed.

Hermitian codes offer desirable properties such as large code lengths, good error-correction at high code rates, etc. The main problem in making Hermitian codes practical is to find a way of performing the required computations in a fast and memory efficient way so as to satisfy resource and throughput constraints imposed by the systems. The paper presents some architecture for Hermitian decoders which enhance their applicability in communication systems. Formulae and architectures for gap detection and address generation unit for satisfying memory constraints have been presented, which amount to 50% savings in storage area and 10% savings in the number of clock cycles reported in literature. A semi-parallel architecture is proposed as a solution to the latency and resource requirements tradeoff, which improves the throughput about q times compared to the word-serial architecture at an expense of some q times more adders, multipliers and simple multiplexers, where the code is defined over GF(q2). For a t error correcting code, the resource load of the parallel architectures is about gamma(t/q + (q-3)/4)( t/q + (q-3)/4 + 1) times this architecture, where gamma is the resource requirement ratio of a multiplier and an inverter

The implementation of the Water Framework Directive (WFD) across the EU, and the growing international emphasis on the management of water quality is giving rise to an expanding market for novel, miniaturized, intelligent monitoring systems for freshwater catchments, transitional and coastal waters. This paper describes the "SmartCoast" multi sensor system for water quality monitoring. This system is aimed at providing a platform capable of meeting the monitoring requirements of the Water Framework Directive. The key parameters under investigation include temperature, phosphate, dissolved oxygen, conductivity, pH, turbidity and water level. The "plug and play" capabilities enabled by the wireless sensor network (WSN) platform developed at Tyndall allow for integration of sensors as required are described, as well as the custom sensors under development within the project.

Abstract—Wireless sensor networks are collections of autonomous devices with computational, sensing and wireless communication capabilities. Research in this area has been growing in the past few years given the wide range of applications that can benefit from such a technology. This paper reports on a joint project between The Tyndall National Institute and the Computer Science Department at University College Cork, Ireland in developing a novel miniaturised modular platform for wireless sensor networks. The system architecture, hardware and software will be discussed as well as details of the deployment scenario chosen for the project – a car park management system. Results and problems encountered during deployment will be presented Keywords-component Wireless Sensor Networks, deployment, car park monitoring

A complete wireless sensor network solution for carpark management is presented in this paper. The system architecture and design are first detailed, followed by a description of the current working implementation, which is based on our DSYS25z sensing nodes. Results of a series of real experimental tests regarding connectivity, sensing and network performance are then discussed. The analysis of link characteristics in the car-park scenario shows unexpected reliability patterns which have a strong influence on MAC and routing protocol design. Two unexpected link reliability patterns are identified and documented. First, the presence of the objects (cars) being sensed can cause significant interference and degradation in communication performance. Second, link quality has a high temporal correlation but a low spatial correlation. From these observations we conclude that a) the construction and maintenance of a fixed topology is not useful and b) spatial rather than temporal message replicates can improve transport reliability

Wireless sensor networks are collections of autonomous devices with computational, sensing and wireless communication capabilities. Research in this area has been growing in the past few years given the wide range of applications that can benefit from such a technology. In this paper, the development of a highly modular and miniaturized wireless platform for sensor networks is described. The system incorporates a radio transceiver (in the 2.4 GHz ISM Band) with embedded protocol software to minimize power consumption and maximize data throughput. Additional input capability for sensor and actuator integration can be incorporated seamlessly due to the modular nature of the system. The total system is packaged in a modular 25 mm cubed form factor. A smaller, (10 mm cubed), prototype is currently under development. Ongoing development of highly miniaturized nodes is discussed.

In this demonstration, a new sensor platform named DSYS25 is presented. The platform has a unique hardware design and runs a customized version of the TinyOS operating system. Transceiver hardware and packaging distinguish the D-Systems platform from other available designs.