Static Sensor Research Articles

Network Optimization of Carbon Monoxide Sensor Nodes in the Metropolitan Region of São Paulo

Air pollution is one of the biggest problems affecting large urban areas. Better monitoring of regions suffering from this type of pollution is in the interest of public health. Although many cities employ sensors to monitor air pollution, a current concern is how to establish the ideal number of sensors to monitor a given geographical region. To address this concern, this research proposes a method to optimize the number of sensors in an air pollution monitoring network to cover a given region efficiently and precisely and uses the metropolitan region of São Paulo, Brazil, and CO sensors as an example. The model of Fragmentation into Groups via Routes is proposed to distribute sensors within micro-regions that display similar air pollution characteristics. A network of virtual sensors is created, and the output of each sensor is established using a method of spatial interpolation called IDW. To identify the optimum sensor configuration, a genetic algorithm is used to assess the topology with the lowest variance of data spread. A lesser number of sensor stations to be treated leads to faster responses to sudden changes in urban conditions. Therefore, municipality authorities can take quick measures to improve the population’s wellness.

Examination of Thinning Inspection Method for Ferromagnetic Steel Tubes Using Velocity Effect of Static Magnetic Field

The high-speed movement of an electromagnetic sensor comprising static excitation and detection coils in a steel tube generates an eddy current in the tube. The presence of a defect in the steel tube may be detected via changes in the eddy current. In general, the ferromagnetic heat transfer tubes in oil and chemical plants are inspected for defects by moving an internally inserted electromagnetic sensor at a high speed of 1 m/s. Although previous studies have focused on the detection of localized thinning in the steel tube, there are no reports on a wide range of thinning in the steel tube. Therefore, this study investigated a thickness measurement method for a wide range of thinning in the steel tube using only static magnetic field. Consequently, the possibility of measuring the thickness in the steel tube based on the flux density inside the detection coil of the proposed static magnetic inspection sensor moving at 1 m/s was demonstrated. The flux density in the steel tube was evaluated by the three-dimensional nonlinear finite element method considering the velocity effect.

Investigation of personal air pollution exposures and occupants’ fresh air demands in two office buildings in Switzerland

This study assessed personal exposures to indoor air pollutants in office environments with dynamic occupancy profiles. A two-week field campaign was conducted in two Swiss office buildings, examining three target areas in each building. We measured air temperature (T), relative humidity (RH), CO2, PM2.5, and PM10 levels at three stationary locations of each target area. Four office workers carried a customized vest to monitor personal CO2 and PM levels near their breathing zone. Using smartwatches, the workers completed hourly surveys about their activity profiles and preferences for outdoor air. We found that CO2 and PM10 concentrations were 1.2–1.3x and 1.8–2.5x higher, respectively, in the personal monitor compared to stationary sensors. Occupants had higher personal CO2 levels during sitting activity and higher personal PM levels during standing activity. Lunch/coffee/call activities were associated with discernible peaks of personal exposure to CO2, PM2.5, and PM10. The higher T coupled with lower RH measured at the personal level was associated with higher occupant demands for outdoor air.

Urban green spaces and variation in cooling in the humid tropics: The case of Paramaribo

The urban climate affects more than half the world’s population, and urban green spaces are considered a nature-based solution to alleviate the urban heat island effect and adapt cities to climate change. Knowledge on urban green spaces cooling draws mostly on data from temperate climates, and similar research in humid tropical climates often focuses on the dry season. This study presents year-round temperature and humidity data for sixteen stationary sensors in Paramaribo, the capital of Suriname, and remotely sensed land surface temperatures for these locations. Analysis was done of diurnal and seasonal dynamics, the extent of urban green space cooling and the relation between locational characteristics and the micro-climate. Results show cooling up to 2.5 ºC with distinct seasonal patterns, and that locations exhibiting stronger cooling during the day have smaller temperature ranges of about 4 ºC at night compared to ranges of 5–7 ºC at other locations. Locations with more trees and complex vegetation structure have temperatures that are 1–5 ºC lower than other locations, but this cooling decreases when the ratio with impervious surfaces increases. Land surface temperature differences between more vegetated and built-up areas reach up to 2.5 ºC. High correlations found among micro-climate indicators imply easier comparison between studies when using any of these indicators, even if not the same ones. The longer term data collected in our study enables investigating urban green space cooling taking into account seasonality typical to the humid-tropics and finds that this cooling can help cities in the Caribbean region adapt to temperature extremes, despite high humidity. Our study further provides an example for overcoming data scarcity and contributes to developing strategies for mitigating increasing heat-related health risks in the humid tropics.

Probabilistic indoor tracking of Bluetooth Low-Energy beacons

We construct a practical and real-time probabilistic framework for fine target tracking. In our scenario, a Bluetooth Low-Energy (BLE) device navigating in the environment publishes BLE packets that are captured by stationary BLE sensors. The aim is to accurately estimate the live position of the BLE device emitting these packets. The framework is built upon a hidden Markov model (HMM), the components of which are determined with a combination of heuristic and data-driven approaches. In the data-driven part, we rely on the fingerprints formed priorly by extracting received signal strength indicators (RSSI) from the packets. These data are then transformed into probabilistic radio-frequency maps that are used for measuring the likelihood between an RSSI data and a position. The heuristic part involves the movement of the tracked object. Having no access to any inertial information of the object, this movement is modeled with Gaussian densities with variable model parameters that are to be determined heuristically. The practicality of the framework comes from the associated small parameter set used to discretize the components of the HMM. By tuning these parameters, such as the grid size of the area, the mask size and the covariance of the Gaussian; a probabilistic filtering becomes tractable for discrete state spaces. The filtering is then performed by the forward algorithm given the instantaneous sequential RSSI measurements. The performance of the system is evaluated by taking the mean squared errors of the most probable positions at each time step to their corresponding ground-truth positions. We report the statistics of the error distributions and see that we achieve promising results. The approach is also finally evaluated by its runtime and memory usage.

Mesoscale soil moisture survey by mobile cosmic-ray neutron sensor across various landscapes in the Heihe River Basin

The cosmic-ray neutron technology is a novel method for upscaling soil moisture, holding significant importance in drought monitoring, soil water storage estimation, calibration of remote sensing products, and data assimilation of land surface models. However, most studies conducted soil moisture measurement experiments using static Cosmic-Ray Neutron Sensors (CRNS) and are often limited to a single landscape, lacking in-situ measurement for the cosmic-ray neutron rover across various landscapes. This study carried out soil moisture observation experiments using a cosmic-ray neutron rover in nine grassland plots, nine farmland plots, and nine desert plots of the Heihe River Basin, Northwest China. The neutron counts displayed clear variations among different plots, and an equation could represent the relationship between neutron intensity and multiple variables near the ground. The parameter N0 was the smallest for grassland and greatest for desert, and there was a significant negative correlation between Normalized Difference Vegetation Index and N0 (P <0.05). Furthermore, the performance of CRNS notably improved after considering soil lattice water and soil organic matter. The Root Mean Square Error (RMSE) between the measured and oven-dried soil moisture increased with radius. Soil moisture measured by CRNS strongly supported the 100 m diameter footprint assumption (RMSE <0.044 g/g). The study offers valuable insights into the application and promotion of cosmic-ray neutron technology across various landscapes within a basin.

A Flexible Implant for Multi-Day Monitoring of Colon Segment Activity.

Monitoring of colon activity is currently limited to tethered systems like anorectal manometry. These systems have significant drawbacks, but fundamentally limit the observation time of colon activity, reducing the likelihood of detecting specific clinical events. While significant technological advancement has been directed to mobile sensor capsules, this work describes the development and feasibility of a stationary sensor for describing the coordinated activity between neighboring segments of the colon. Unlike wireless capsules, this device remains in position and measures propagating pressure waves and impedances between colon segments to describe activity and motility. This low-power, flexible, wireless sensor-the colon monitor to capture activity (ColoMOCA) was validated in situ and in vivo over seven days of implantation. The ColoMOCA diameter was similar to common endoscopes to allow for minimally invasive diagnostic placement. The ColoMOCA included two pressure sensors, and three impedance-sensing electrodes arranged to describe the differential pressures and motility between adjacent colon segments. To prevent damage after placement in the colon, the ColoMOCA was fabricated with a flexible polyimide circuit board and a silicone rubber housing. The resulting device was highly flexible and suitable for surgical attachment to the colon wall. In vivo testing performed in eleven animals demonstrated suitability of both short term (less than 3 hours) and 7-day implantations. Data collected wirelessly from animal experiments demonstrated the ColoMOCA described colon activity similarly to wired catheters and allowed untethered, conscious monitoring of organ behavior.

End-to end decision support system for sleep apnea detection and Apnea-Hypopnea Index calculation using hybrid feature vector and Machine learning

Sleep apnea is a disease that occurs due to the decrease in oxygen saturation in the blood and directly affects people's lives. Detection of sleep apnea is crucial for assessing sleep quality. It is also an important parameter in the diagnosis of various other diseases (diabetes, chronic kidney disease, depression, and cardiological diseases). Recent studies show that detection of sleep apnea can be done via signal processing, especially EEG and ECG signals. However, the detection accuracy needs to be improved. In this paper, a ML model is used for the detection of sleep apnea using 19 static sensor data and 2 dynamic data (Sleep score and Arousal). The sensor data is recorded as a discrete signal and the sleep process is divided into 4.8 M segments. In this work, 19 different sensor data sets were recorded with polysomnography (PSG). These data sets have been used to perform sleep scoring. Then, arousal status marking is done. Model training was carried out with the feature vector consisting of 21 data obtained. Tests were performed with eight different machine learning techniques on a unique dataset consisting of 113 patients. After all, it was automatically determined whether people were diseased (a kind of apnea) or healthy. The proposed model had an average accuracy of 97.27%, while the recall, precision, and f-score values were 99.18%, 95.32%, and 97.20%, respectively. After all, the model that less feature engineering, less complex classification model, higher dataset usage, and higher classification performance has been revealed.

Mobile Structural Health Monitoring Based on Legged Robots

With the advancements in information, communication, and sensing technologies, structural health monitoring (SHM) has matured into a substantial pillar of infrastructure maintenance. In particular, wireless sensor networks have gradually been incorporated into SHM, leveraging new opportunities towards reduced installation efforts and enhanced flexibility and scalability, as compared to cable-based SHM systems. However, wireless sensor nodes are installed at fixed locations and need to be employed at high density to reliably monitor large infrastructure, which may cause high installation costs. Furthermore, the limited power autonomy of wireless sensor networks, installed at fixed locations for unattended long-term operation, still represents a significant constraint when deploying stationary wireless sensor nodes for SHM. To resolve the critical constraints stemming from costly high-density deployment and limited power autonomy, a mobile structural health monitoring concept based on legged robots is proposed in the study reported in this paper. The study explores the accuracy and cost-efficiency of deploying legged robots in dense measurement setups for wireless SHM of civil infrastructure, aiming to gain insights into the advantages of mobile wireless sensor nodes in general and of legged robots in particular, in terms of obtaining rich information on the structural condition. As is shown in this paper, the legged robots, as compared to stationary wireless sensor nodes, require a smaller number of nodes to be deployed in civil infrastructure to achieve rich sensor information, entailing more cost-efficient, yet accurate, SHM. In conclusion, this study represents a first step towards autonomous robotic fleets advancing structural health monitoring.

Investigation of indicators for personal exposure and occupancy in offices by using smart sensors

Despite the advent of smart building sensors, real-time methods for capturing inhalation exposure and occupancy are limited. This study utilized stationary and wearable environmental sensors, along with Decision Tree and correlation analyses, to determine the optimal set of indicators for approximating dynamic inhalation exposures and occupancy in office environments. In a 2-week field campaign in two Swiss office buildings, real-time measures of air temperature, relative humidity, CO2, and size-resolved particle concentrations were taken at two scales: personal – via personalized vests with sensors for personal exposure detection; and room – via stationary sensors at sidewalls and office desks. Occupancy and activity profiles were collected at three scales: personal – via smartwatches; room – via visual inspections; and building – via a cloud-based location service system. A desk-mounted CO2 sensor in the center of an office was an effective indicator of personal exposures to CO2, PM2.5, and PM10. Sidewall CO2 measurements accurately captured room occupancy in open-plan offices, while desk-mounted PM10 sensors were effective in areas with transient occupancy (e.g., cafeteria). Findings from this study aid improved understanding of the complex dynamics of air pollutants in offices, and could support the development of refined methods for smart building monitoring and ventilation control.

Real-Time 3D Reconstruction Pipeline for Room-Scale, Immersive, Medical Teleconsultation

Medical teleconsultation was among the initial use cases for early telepresence research projects since medical treatment often requires timely intervention by highly specialized experts. When remote medical experts support interventions, a holistic view of the surgical site can increase situation awareness and improve team communication. A possible solution is the concept of immersive telepresence, where remote users virtually join the operating theater that is transmitted based on a real-time reconstruction of the local site. Enabled by the availability of RGB-D sensors and sufficient computing capability, it becomes possible to capture such a site in real time using multiple stationary sensors. The 3D reconstruction and simplification of textured surface meshes from the point clouds of a dynamic scene in real time is challenging and becomes infeasible for increasing capture volumes. This work presents a tightly integrated, stateless 3D reconstruction pipeline for dynamic, room-scale environments that generates simplified surface meshes from multiple RGB-D sensors in real time. Our algorithm operates directly on the fused, voxelized point cloud instead of populating signed-distance volumes per frame and using a marching cube variant for surface reconstruction. We extend the formulation of the dual contouring algorithm to work for point cloud data stored in an octree and interleave a vertex-clustering-based simplification before extracting the surface geometry. Our 3D reconstruction pipeline can perform a live reconstruction of six incoming depth videos at their native frame rate of 30 frames per second, enabling the reconstruction of smooth movement. Arbitrarily complex scene changes are possible since we do not store persistent information between frames. In terms of mesh quality and hole filling, our method falls between the direct mesh reconstruction and expensive global fitting of implicit functions.

Adaptive Techniques for the Online Estimation of Spatial Fields With Mobile Sensors

The regression-based reconstruction of spatial fields by static sensors requires the number of measurements be at least equal to the number of unknown parameters in order to provide a unique solution. Improvements include the use of mobile sensor networks to arrive at higher accuracy estimates. An alternate approach utilizes adaptive techniques to update the parameter estimates online. Such adaptive techniques need persistence of excitation to ensure convergence of the parameter and function estimates. It is shown that the motion of just a single sensor is a necessary condition for persistence of excitation and, subsequently, of function convergence. Having a prescribed sensor guidance is also a sufficient condition for function convergence. First, a Lyapunov-based adaptation and guidance is presented and it is shown that persistence of excitation necessitates the time dependence of the outer product of the regressor vector with itself and which in turn requires the motion of the sensor within the spatial domain. Subsequently, a gradient-based adaptive scheme relaxes the Lyapunov-based sensor guidance to any user-defined guidance as long as the outer product of the associated rank-one regressor vector with itself has a uniformly positive definite integral. The latter adaptation is conducive to the inclusion of platform dynamics resulting in a vehicle control design needed to attain parameter convergence. Extension of the centralized solution to distributed estimation using a network of mobile sensors is also examined. These results are demonstrated with simulations in one and two dimensions.

Validation of a neural network based model to predict foF2

This work presents the application of Neural Networks (NNs) techniques to develop an empirical model for foF2, the critical frequency of the F2 ionospheric layer. Prediction of foF2 parameter is important because of the paucity of data in some areas of the globe (due to uneven distributions of ground- based measuring equipment) and, also, because of the roles it plays in high frequency communications and navigation purposes. Hourly values of foF2 corresponding to the period 1985–2007 and obtained from 105 World-wide distributed ionospheric stations, have been considered in this work. A statistical analysis of the difference between NN model-derived and the corresponding experimental foF2 values was carried out for eleven (11) selected ionospheric stations during periods of very high (year 2000), relatively low (year 2006) and very low (year 2008) solar activity. In particular, the ionospheric behavior during the years 2000 and 2008 appeared to be very challenging to be captured for any empirical model. For comparison purposes, the same kind of statistical analysis was performed with foF2 values as obtained from the International Reference Ionosphere (IRI) model and from NeQuick 2 model. Results of the statistical comparison confirmed that the neural network modelling approach is suitable to describe in a climatological way the main characteristics of the ionosphere F2 layer peak frequency, also over the extended geographic region where the additional sensor stations were located. The results obtained also indicate that the NN predictions and NeQuick 2 algorithm driven by solar flux (F10.7) index have similar performance.

An Enhanced Food Digestion Algorithm for Mobile Sensor Localization.

Mobile sensors can extend the range of monitoring and overcome static sensors' limitations and are increasingly used in real-life applications. Since there can be significant errors in mobile sensor localization using the Monte Carlo Localization (MCL), this paper improves the food digestion algorithm (FDA). This paper applies the improved algorithm to the mobile sensor localization problem to reduce localization errors and improve localization accuracy. Firstly, this paper proposes three inter-group communication strategies to speed up the convergence of the algorithm based on the topology that exists between groups. Finally, the improved algorithm is applied to the mobile sensor localization problem, reducing the localization error and achieving good localization results.

Real-Time Embedded Eye Image Defocus Estimation for Iris Biometrics.

One of the main challenges faced by iris recognition systems is to be able to work with people in motion, where the sensor is at an increasing distance (more than 1 m) from the person. The ultimate goal is to make the system less and less intrusive and require less cooperation from the person. When this scenario is implemented using a single static sensor, it will be necessary for the sensor to have a wide field of view and for the system to process a large number of frames per second (fps). In such a scenario, many of the captured eye images will not have adequate quality (contrast or resolution). This paper describes the implementation in an MPSoC (multiprocessor system-on-chip) of an eye image detection system that integrates, in the programmable logic (PL) part, a functional block to evaluate the level of defocus blur of the captured images. In this way, the system will be able to discard images that do not have the required focus quality in the subsequent processing steps. The proposals were successfully designed using Vitis High Level Synthesis (VHLS) and integrated into an eye detection framework capable of processing over 57 fps working with a 16 Mpixel sensor. Using, for validation, an extended version of the CASIA-Iris-distance V4 database, the experimental evaluation shows that the proposed framework is able to successfully discard unfocused eye images. But what is more relevant is that, in a real implementation, this proposal allows discarding up to 97% of out-of-focus eye images, which will not have to be processed by the segmentation and normalised iris pattern extraction blocks.

Real-time biodiversity analysis using deep-learning algorithms on mobile robotic platforms.

Ecological biodiversity is declining at an unprecedented rate. To combat such irreversible changes in natural ecosystems, biodiversity conservation initiatives are being conducted globally. However, the lack of a feasible methodology to quantify biodiversity in real-time and investigate population dynamics in spatiotemporal scales prevents the use of ecological data in environmental planning. Traditionally, ecological studies rely on the census of an animal population by the "capture, mark and recapture" technique. In this technique, human field workers manually count, tag and observe tagged individuals, making it time-consuming, expensive, and cumbersome to patrol the entire area. Recent research has also demonstrated the potential for inexpensive and accessible sensors for ecological data monitoring. However, stationary sensors collect localised data which is highly specific on the placement of the setup. In this research, we propose the methodology for biodiversity monitoring utilising state-of-the-art deep learning (DL) methods operating in real-time on sample payloads of mobile robots. Such trained DL algorithms demonstrate a mean average precision (mAP) of 90.51% in an average inference time of 67.62 milliseconds within 6,000 training epochs. We claim that the use of such mobile platform setups inferring real-time ecological data can help us achieve our goal of quick and effective biodiversity surveys. An experimental test payload is fabricated, and online as well as offline field surveys are conducted, validating the proposed methodology for species identification that can be further extended to geo-localisation of flora and fauna in any ecosystem.

Enabling fire source localization in building fire emergencies with a machine learning-based inverse modeling approach

During building fire emergency response (ER), one of the most important tasks for firefighting is to locate the fire source. Currently, incident commanders on the fire ground determine the origin of fire sources, relying mostly on the reports of firefighters entering the building or the information from on-site witnesses. This inevitably requires firefighters to approach the fire origin and observe, causing significant dangers and risks. To overcome this issue, an inverse model for building fire source localization is developed. A machine learning (ML)-based inverse modeling approach is used to build the complex relationship between the fire source location and on-site temperature sensor measurements, so that an inversion of the fire source based on the temperature observations can be enabled. By taking fire in an actual building floor as a case study, the inverse model for fire source localization using simulated fire temperature data is established. ML methods of feed-forward neural networks (FFNNs) and long short-term memory (LSTM) are respectively employed. Results indicate that LSTM in stateless mode is more appropriate to establish the relative models because more accurate recognition of the fire source room can be achieved during the initial and growth stages of fire development. Besides, by considering different means of on-site data acquisition in building fire emergencies, fire source localization via stationary temperature sensor arrays and portable temperature measuring devices is realized. The formulated models achieve recognition accuracy of 99.5% (direct recognition) and 92.1% (in top-3 candidates) in the first 15 min of the fires, respectively. The effectiveness of using the ML-based inverse modeling technique for fire source localization in building fire emergencies is demonstrated.

Intrusion Detection Mechanism for Empowered Intruders Using IDEI

In the past, intrusion detection has been extensively investigated as a means of ensuring the security of wireless sensor networks. Anti-recon technology has made it possible for an attacker to get knowledge about the detecting nodes and plot a route around them in order to evade detection. An "empowered intruder" is one who poses new threats to current intrusion detection technologies. Furthermore, the intended impact of detection may not be obtained in certain subareas owing to gaps in coverage caused by the initial deployment of detection nodes at random. A vehicle collaboration sensing network model is proposed to solve these difficulties, in which mobile sensing cars and static sensor nodes work together to identify intrusions by empowered intruders. An algorithm for mobile sensing vehicles, called Intrusion Detection Mechanism for Empowered Intruders(IDEI), and a sleep-scheduling technique for static nodes form the basis of our proposal. Sophisticated intruders will be tracked by mobile sensors, which will fill in the gaps in coverage, while static nodes follow a sleep schedule and will be woken when the intruder is discovered close. Our solution is compared to current techniques like Kinetic Theory Based Mobile Sensor Network (KMsn)and Mean Time to Attacks (MTTA) in terms of intrusion detection performance, energy usage, and sensor node movement distance. IDEI's parameter sensitivity is also examined via comprehensive simulations. It is clear from the theoretical analysis and simulation findings that our idea is more efficient and available.

Flexible energy storage patch based on NiPS3/graphene zinc-ion hybrid supercapacitor for integrated biosensors

The rapidly developing self-powered biosensor technologies require flexible, robust, and safe energy storage devices. One of the key factors of a self-powered system is the integration of energy storage device components with bio-sensors. Here, we developed a hybrid self-powered biosensor system for health monitoring comprising zinc-ion supercapacitor (ZISC) wearable patch, introducing a two-dimensional (2D) NiPS3 as a metal phosphorus chalcogenide capacitive electrode. Due to its inherent 2D nanosheet morphology, NiPS3 electrode confined with graphene (NiPS3@graphene) shows good electrochemical performance. The ZISC printed with NiPS3@graphene and zinc on cellulose substrate followed by hydrogel electrolyte results in a powerful, robust, and flexible device. It exhibits stable energy storage performance, retaining 86% capacity after 1000 charge–discharge cycles and good mechanical flexibility at different bending angles, with a retaining capacity of 97.7% compared to initial performance after 500 bending cycles. Finally, a self-powered biosensor was developed by transforming a thin and lightweight ZISC tandem device on a medical-grade cellulose patch. This patch is demonstrated to power a wearable temperature sensor with a connection via a Bluetooth module and stationary glucose sensor to establish real-time health monitoring. Our results show the significance of 2D NiPS3@graphene-based zinc-ion supercapacitors in an application as an integrated patch for powering health monitoring systems.

Multiagent Deep Reinforcement Learning With Demonstration Cloning for Target Localization

In target localization applications, readings from multiple sensing agents are processed to identify a target location. The localization systems using stationary sensors use data fusion methods to estimate the target location, whereas other systems use mobile sensing agents (UAVs, robots) to search the area for the target. However, such methods are designed for specific environments, and hence are deemed infeasible if the environment changes. For instance, the presence of walls increases the environment’s complexity and affects the collected readings and the mobility of the agents. Recent works explored Deep Reinforcement Learning (DRL) as an efficient and adaptable approach to tackle the target search problem. However, such methods are either designed for single-agent systems or for non-complex environments. This work proposes two novel Multi-Agent DRL models for target localization through search in complex environments. The first model utilizes Proximal Policy Optimization, Convolutional Neural Networks, Convolutional AutoEncoders to create embeddings, and a shaped reward function using Breadth First Search to obtain cooperative agents that achieve fast localization at low cost. The second model improves the first model in terms of computational complexity by replacing the shaped reward with a simple sparse reward, subject to the availability of Expert Demonstrations. Expert demonstrations are used in Demonstration Cloning, a novel method that utilizes demonstrations to guide the learning of new agents. The proposed models are tested on a scenario of radioactive target localization, and benchmarked with existing methods, showing efficacy in terms of localization time and cost, in addition to learning speed and stability.