Target tracking is challenging when sensor detection thresholds cause state-dependent missed detections, particularly in multi-sensor scenarios with clutter and uncertain target existence. A recently developed missed detection framework models detection probability as a function of target state, sensor characteristics, and detection threshold, but it is limited to individual measurements and does not address the recursive tracking problem. This work extends the framework using a Bernoulli filter formulation to jointly handle recursive target tracking, clutter, and target existence uncertainty. A Bernoulli particle filter is evaluated in a simulated 2D multi-sensor tracking scenario with nonlinear measurements, clutter, and detection uncertainty. Incorporating accurate detection threshold knowledge reduces the generalized optimal subpattern assignment (GOSPA) metric by 62.4

The marginalized particle filter (MPF) is known to often outperform the standard particle filter (PF) in terms of estimation accuracy for the same number of particles. This is due to the fact that the MPF uses a Kalman filter (KF) to handle a linear sub-part of the system in a more efficient way than a particle approximation. However, for certain systems, marginalization implies no gain (MING), resulting in identical output distribution, including estimation accuracy. This article explains when and why this happens in two steps. The first step is to show that the MPF and PF algorithms are algebraically identical in the case the KF covariance is zero. The second step is to show that for MING systems the KF in the MPF degenerates, in that the KF covariance approaches zero. We provide a practical guide for easily checking if a system is a MING system or not, and show by examples that MING systems exist, and in fact are quite common in, e.g., tracking applications.

This paper introduces a signal strength-based direction of arrival (DOA) estimation approach for directional sensors that explicitly accounts for missed detections. In traditional phase-based DOA estimation frameworks, negative information from expected emitters that fall below the detection threshold fall outside the scope of standard measurement models. Unlike phase-based DOA estimation methods, the proposed approach relies only on received signal strength measurements. As a result, missed detections arise naturally from the sensing and detection process and convey valuable information via the known detection thresholds. By incorporating both detected signals and missed detections into the likelihood function, we develop a probabilistic estimation method that fully leverages the underlying measurement and detection models. Simulation results show that the proposed method significantly improves DOA estimation accuracy compared to baseline techniques, particularly in challenging scenarios with high missed-detection rates. Real-world experiments using Bluetooth Low Energy (BLE) signals and directional antennas further validate the effectiveness of the approach, demonstrating substantial performance gains. These findings highlight the value of modeling missed detections in sensor array processing and open new avenues for enhancing localization performance in wireless communication systems.

Camera traps have become a common tool for wildlife monitoring efforts in ecological research and biodiversity conservation. Wildlife classification models have benefited from the increase in wildlife visual data. These models reach high levels of accuracy on curated, high-quality datasets. However, their performance remains sensitive to real-world environmental constraints. They often produce inconsistent predictions when performing inference on temporally coherent sequences. The predicted label for a single individual shifts rapidly between frames. This study exploits the temporal nature of camera-trap data to augment inferred predictions from a wildlife classification model. Specifically, we adopt several standard Multi-Object Tracking (MOT) models to link detections across consecutive frames. The curated trajectories are used to fuse the softmax class probabilities. The fused probability score produces a single consensus class label estimate that overrides misclassifications caused by noise. The analysis of the experimental results shows that our proposed strategy improves over a standalone classifier over all datasets and for each metric. Specifically, the best-performing MOT models gain a weighted F1-Score of 5.1

Fine-grained wildlife re-identification remains a challenging area in research. Current state-of-the-art approaches apply a detection and re-identification pipeline. We propose a one-stage end-to-end detection and re-identification model that performs identity searching within the latent space. We adopt DINOv2 for robust spatial geometry and MegaDescriptor for wildlife re-identification. We enhance latent queries with prompt re-identification features. A detection decoder queries the scene latent space to establish object boundaries around the target identity. Preliminary findings reflect a competitive mean average precision score of 30.584

Direction of arrival (DOA) estimation is a fundamental problem in signal processing and has applications in various fields such as radar, sonar, and acoustics. In this article, we propose a method for DOA estimation using the received power at each sensor. The method is based on the directional sensitivity of the sensor elements at various frequencies. We model the directional sensitivity using a Fourier series (FS) model, where the parametric model enables Cram & eacute;r-Rao lower-bound (CRLB) computations. The FS model is estimated from measurements of a wideband noise signal. To estimate the DOA, the measured power profile is compared to the FS model using the least-squares (LS) method. The proposed power-based method offers several advantages over classical time-delay methods, particularly in allowing arbitrarily small arrays and still handling broadband signals. Additionally, it enables low-rate sampling, which simplifies hardware requirements and significantly reduces processor load. In numerical evaluations with a microphone array and natural sound sources, we still benchmark our method against state-of-the-art time-delay methods. Real-world experiments show promising results, performing on par with the best of the other evaluated methods for all natural signals, despite relying on significantly less information. A key benefit is robustness against array size limitations. By utilizing the received signal power instead of time delays or phase information, the method enables small arrays with great DOA resolution. Furthermore, outdoor data collected a year after calibration confirms its robustness, demonstrating consistent performance over time.

Road roughness significantly affects vehicle vibrations and ride quality. We introduce a Kalman filter (KF)-based method for estimating road roughness in terms of the international roughness index (IRI) by fusing inertial and speed measurements, offering a cost-effective solution for pavement monitoring. The method involves system identification on a physical vehicle to estimate realistic model parameters, followed by KF-based reconstruction of the longitudinal road profile to compute IRI values. It explores IRI estimation using vertical and lateral vibrations, the latter more common in modern vehicles. Validation on 230 km of real-world data shows promising results, with IRI estimation errors ranging from 1% to 10% of the reference values. However, accuracy deteriorates significantly when using only lateral vibrations, highlighting their limitations. These findings demonstrate the potential of KF-based estimation for efficient road roughness monitoring.

Mitigating human-wildlife conflict seeks to resolve unwanted encounters between these parties. Computer Vision provides a solution to identifying individuals that might escalate into conflict, such as members of the Big Five African animals. However, environments often contain several varied species. The current state-of-the-art animal classification models are trained under a closed-world assumption. They almost always remain overconfident in their predictions even when presented with unknown classes. This study investigates out-of-distribution (OOD) detection of wildlife, specifically the Big Five. To this end, we select a parametric Nearest Class Mean (NCM) and a non-parametric contrastive learning approach as baselines to take advantage of pretrained and projected features from popular classification encoders. Moreover, we compare our baselines to various common OOD methods in the literature. The results show feature-based methods reflect stronger generalisation capability across varying classification thresholds. Specifically, NCM with ImageNet pre-trained features achieves a 2

Current state-of-the-art Wildlife classification models are trained under the closed world setting. When exposed to unknown classes, they remain overconfident in their predictions. Open-set Recognition (OSR) aims to classify known classes while rejecting unknown samples. Several OSR methods have been proposed to model the closed-set distribution by observing the feature, logit, or softmax probability space. A significant drawback of many existing approaches is the requirement to retrain the pre-trained classification model with the OSR-specific strategy. This study contributes a post-processing OSR method that measures the agreement between the models’ features and predicted logits. We propose a probability distribution based on an input’s distance to its Nearest Class Mean (NCM). The NCM-based distribution is then compared with the softmax probabilities from the logit space to measure agreement between the NCM and the classification head. Our proposed strategy ranks within the top three on two evaluated datasets, showing consistent performance across the two datasets. In contrast, current state-of-the-art methods excel on a single dataset. We achieve an AUROC of 93.41 and 95.35 for African and Swedish animals. The code will be released publicly upon acceptance of this paper.

Due to recent events that have demonstrated the vulnerabilities of global navigation satellite systems (GNSS) there has been an increased interest in alternative methods for localization. One traditional alternative method is terrain-aided navigation (TAN), where a platform localizes itself by measuring the terrain elevation and comparing it to a digital elevation map (DEM). While single-agent TAN has been extensively studied, multi-agent TAN remains less explored. This paper addresses the multi-agent TAN problem with a focus on its properties. We formulate a weighted least squares (WLS) estimator for computing a snapshot solution to the problem and formulate a Cramer-Rao Lower Bound (CRLB) to evaluate it. Using the expressions for the estimator and the CRLB we are able to highlight some insightful properties of the problem. The findings are verified in a simulation study where we evaluate the performance with respect to the altitude sensor accuracy, the group formation accuracy, the number of agents and their formation. Notably, we observe that the solution is relatively insensitive to errors in agent position, suggesting that low-accuracy inertial navigation systems and distance sensors are sufficient for determining their positions. Increasing the number of agents beyond a few seems to have a large effect on both the efficiency and robustness of the estimator, which lessens as the number of agents increases. However, increasing the number of agents does not compensate for poor altitude sensor quality. Additionally, while spatial separation between agents is important for effective map utilization, further separation beyond a certain point does not enhance performance. These findings provide design guidelines for multi-agent TAN systems and identify areas for further research.

As human settlement expands into the natural habitats of wild animals, the conflicts between humans and wildlife increases. The human-elephant conflict causes a tremendous amount of damage, often to poor villages close to the savannah. In this paper, we continue our earlier reported research on a geophone network aimed for elephant localisation by focusing on the detection challenge. We have now collected larger sets of seismic data with footsteps from both elephants and other big animals including humans. To detect the footsteps, a method is developed that analyses features of the geophone signal, which are then compared to those of an elephant footstep. The method detects 54% of the footsteps and has a classification accuracy of 89 %. Subsequently, the detected elephant footstep is used to calculate the direction of arrival (DOA) angle using a delay-and-sum beamformer. The direction to an elephant is estimated with good precision on distances ranging from 8 to 30 meters. This research, not only, showcases a practical solution for mitigating human-elephant conflicts, but also underscores the potential of seismic technology in wildlife management and conservation efforts.

This letter proposes a new method for joint state and parameter estimation in uncertain dynamical systems. We exploit the partial errors-in-variables (PEIV) principle and formulate a regression problem in the sense of weighted total least squares, where the uncertainty in the parameter prior is explicitly considered. Based thereon, the PEIV regression can be solved iteratively through the Kalman smoothing and the regularized least squares for estimating the state and the parameter, respectively. The simulations demonstrate improved accuracy of the proposed method compared to existing approaches, including the joint maximum a posterior-maximum likelihood, the expectation maximisation, and the augmented state extended Kalman smoother.

Classifiers based on neural networks (NN) often lack a measure of uncertainty in the predicted class. We propose a method to estimate the probability mass function (PMF) of the different classes, as well as the covariance of the estimated PMF. First, a local linear approach is used during the training phase to recursively compute the covariance of the parameters in the NN. Secondly, in the classification phase another local linear approach is used to propagate the covariance of the learned NN parameters to the uncertainty in the output of the last layer of the NN. This allows for an efficient Monte Carlo (MC) approach for: (i) estimating the PMF; (ii) calculating the covariance of the estimated PMF; and (iii) proper risk assessment and fusion of multiple classifiers. Two classical image classification tasks, i.e., MNIST, and CFAR10, are used to demonstrate the efficiency the proposed method.

We aimed to use the digital platform maintained by the local health service providers in Southeast Sweden for integrated monitoring of disparities in vaccination and morbidity during the COVID-19 pandemic. The monitoring was performed in the adult population of two counties (n = 657,926) between 1 February 2020 and 15 February 2022. The disparities monitored were relocated (internationally displaced), substance users, and suffering from a psychotic disorder. The outcomes monitored were COVID-19 vaccination, SARS-CoV-2 test results, and hospitalization with COVID-19. Relocated residents displayed an increased likelihood of remaining unvaccinated and a decreased likelihood of testing as well as increased risks of primary SARS-CoV-2 infection and hospitalization compared with the general population. Suffering from a major psychiatric disease was associated with an increased risk of remaining unvaccinated and an increased risk of hospitalization but a decreased risk of SARS-CoV-2 infection. From the digital monitoring, we concluded that the relocated minority received insufficient protection during the pandemic, suggesting the necessity for comprehensive promotion of overall social integration. Persons with major psychiatric diseases underused vaccination, while they benefitted from proactively provided testing, implying a need for active encouragement of vaccination. Further research is warranted on legal and ethical frameworks for digital monitoring in vaccination programs.

This paper considers the problem of detecting and tracking objects in a sequence of images. The problem is formulated in a filtering framework, using the output of object-detection algorithms as measurements. An extension to the filtering formulation is proposed that incorporates class information from the previous frame to robustify the classification, even if the object-detection algorithm outputs an incorrect prediction. Further, the properties of the object-detection algorithm are exploited to quantify the uncertainty of the bounding box detection in each frame. The complete filtering method is evaluated on camera trap images of the four large Swedish carnivores, bear, lynx, wolf, and wolverine. The experiments show that the class tracking formulation leads to a more robust classification.

Manufacturing industries are eager to replace traditional robot manipulators with collaborative robots due to their cost-effectiveness, safety, smaller footprint and intuitive user interfaces. With industrial advancement, cobots are required to be more independent and intelligent to do more complex tasks in collaboration with humans. Therefore, to effectively detect the presence of humans/obstacles in the surroundings, cobots must use different sensing modalities, both internal and external. This paper presents a detailed review of sensor technologies used for detecting a human operator in the robotic manipulator environment. An overview of different sensors installed locations, the manipulator details and the main algorithms used to detect the human in the cobot workspace are presented. We summarize existing literature in three categories related to the environment for evaluating sensor performance: entirely simulated, partially simulated and hardware implementation focusing on the ‘hardware implementation’ category where the data and experimental environment are physical rather than virtual. We present how the sensor systems have been used in various use cases and scenarios to aid human–robot collaboration and discuss challenges for future work.

The particle filter (PF) approximates the posterior distribution of the states in filtering problems, and it is well-known that it converges to the true posterior when the number of particles tends to infinity. It would be natural to assume that measures such as mean square error (MSE) decreases monotonically as the number of particles increases. This is, however, not always true. We present a simple two-dimensional linear Gaussian system where the MSE grows initially before it starts to decrease to eventually reach the optimal filter performance, which in this case is provided by the Kalman filter (KF). Other indicators such as the efficient number of particles and trace of the particle covariance show a similar strange behavior. Inspired by this, we derive a condition for what we term projected instability, which means that the particle in the standard SIR PF that gives the best prediction actually increases the state estimation error. For linear systems, this gives an explicit condition in terms of the state space matrices when this situation occurs. Monte Carlo simulations of a large number of random linear systems indicate that everything works as expected as long as the system does not have a projected instability, otherwise the particle filter can perform badly or even diverge.

Reinforcement learning algorithms are commonly used to control dynamical systems with measurable state variables. If the dynamical system is partially observable, reinforcement learning algorithms are modified to compensate for the effect of partial observability. One common approach is to feed a finite history of input-output data instead of the state variable. In this article, we study and quantify the effect of this approach in linear Gaussian systems with quadratic costs. We coin the concept of L-Extra-Sampled-dynamics to formalize the idea of using a finite history of input-output data instead of state and show that this approach increases the average cost.

The Particle filter can in theory estimate the state of any nonlinear system, but in practice it suffers from an exponential complexity in terms of the number of particles as the dimension of the state increases. The marginalized particle filter can potentially reduce this problem by improving the estimates, particularly for lower number of particles. However, it turns out that for certain systems, it does not provide any improvement in the accuracy of the estimate. The core cause of degeneracy is linked to when the uncertainty of the linear state conditioned on the nonlinear state is 0. Conditions for determining when this occurs are presented and applied to common constant velocity, constant acceleration and constant jerk models with various sampling methods. Interestingly, some combinations are useful while others should be avoided. These findings are supported using simulated systems.

The uncertainty in the prediction calculated using the delta method for an over-parameterized (parametric) black-box model is shown to be larger or equal to the uncertainty in the prediction of a canonical (minimal) model. Equality holds if the additional parameters of the overparameterized model do not add flexibility to the model. As a conclusion, for an overparameterized black-box model, the calculated uncertainty in the prediction by the delta method is not underestimated. The results are shown analytically and are validated in a simulation experiment where the relationship between the normalized traction force and the wheel slip of a car is modelled using e.g., a neural network.