We present a novel statistical approach to incorporate uncertainty awareness in model-free distributional deep reinforcement learning for mission and safety-critical robotics. Deep learning predictions are influenced by uncertainties in the data, termed as aleatoric uncertainties, as well as uncertainties in the learning process and model structure, known as epistemic uncertainties. The proposed algorithm, called as Calibrated Evidential Quantile Regression in Deep-Q Networks (CEQR-DQN), addresses key challenges associated with separately estimating aleatoric and epistemic uncertainty in stochastic robotic environments. It combines deep evidential learning with quantile calibration based on the principles of conformal inference to provide explicit, sample-free computations of global uncertainty as opposed to local estimates based on simple variance. Thereby, the proposed approach overcomes limitations of traditional methods in computational and statistical efficiency and handling of out-of-distribution (OOD) observations. Tested on a suite of representative miniaturized Atari games (i.e., MinAtar), CEQR-DQN is shown to surpass similar existing frameworks in scores and learning speed. Its ability to rigorously evaluate uncertainties improves exploration strategies and can serve as a blueprint for other uncertainty-aware robotic algorithms.

Non-destructive inspection of penetrable vertical structures demands innovative investigation paradigms and ad-hoc tools to support timely and efficient maintenance interventions. Although Ground Penetrating Radar (GPR) is an established technology, the inspection of large vertical structures with classical ground-based systems remains challenging when there is limited accessibility, the equipment is difficult to maneuver, the operator’s safety is placed at risk, and the time to perform the survey becomes long. A promising solution to address these limitations suggests the use of mini drones equipped with GPR systems. However, this possibility introduces new challenges in both system design and data processing. A practical approach for drone-based GPR monitoring involves adapting a traditional contact GPR system, combined with appropriate data processing techniques, to perform contactless surveys. This study delves deeper into this approach by comparing the imaging performance achieved when the same ground-coupled GPR system paired with a microwave tomography-based data processing is exploited for contact and contactless surveys. Experimental results related to the inspection of a reinforced concrete wall are presented and discussed.

Preterm neonates are vulnerable to periventricular-intraventricular hemorrhage since the periventricular germinal matrix blood vessels are still immature and fragile until around 36 weeks. While cranial ultrasound imaging (cUS) is the standard technique for diagnosing brain injury in neonates, it has limited accuracy in detecting early-stage hemorrhages due to poor sensitivity and specificity. Transfontanelle thermoacoustic imaging (TTAI) combines the advantages of high contrast from electromagnetic absorption with high-resolution from ultrasound imaging and represents a potential alternative to overcome the limitations of conventional cUS systems. We developed a TTAI system and evaluated its performance in a large animal model of intraventricular hemorrhage (IVH) in vivo. Our system demonstrated the ability to detect hemorrhages as small as 0.1 mL, which is better than the previously reported limit of detection for either ultrasound or photoacoustic imaging in the same animal model. These results indicate that TTAI is a novel imaging modality with high potential for accurate detection of IVH in neonates.

In the expanding landscape of AI-enabled robotics, robust quantification of predictive uncertainties is of great importance. Three-dimensional (3D) object detection, a critical robotics operation, has seen significant advancements; however, the majority of current works focus only on accuracy and ignore uncertainty quantification. Addressing this gap, our novel study integrates the principles of conformal inference (CI) with information theoretic measures to perform lightweight, Monte Carlo-free uncertainty estimation within a multimodal framework. Through a multivariate Gaussian product of the latent variables in a Variational Autoencoder (VAE), features from RGB camera and LiDAR sensor data are fused to improve the prediction accuracy. Normalized mutual information (NMI) is leveraged as a modulator for calibrating uncertainty bounds derived from CI based on a weighted loss function. Our simulation results show an inverse correlation between inherent predictive uncertainty and NMI throughout the model’s training. The framework demonstrates comparable or better performance in KITTI 3D object detection benchmarks to similar methods that are not uncertainty-aware, making it suitable for real-time edge robotics.

Reconfigurable Intelligent Surface (RIS) or Intelligent Reflective Surface (IRS) are sometimes used as synonyms in the literature and refer to the same technology. RIS refers to a material whose physical properties of reflection and transmission can be controlled in order to optimize the path taken by the electromagnetic waves incident upon them. This technology is beginning to gain attention in the context of smart industries and can potentially be a major asset. These surfaces can be manufactured using metasurfaces, which act like arrays of small mirrors that are individually controlled, thus resulting in the capability to reshape the beam incident on the surface to reflect it and redirect it along the desired direction to improve the transmission performance in terms of coverage or quality improvement, such as the bit error rate (BER). With the upcoming 6G, many factories and industries are making the shift towards up to date technologies to make their processes more efficient and more robotized. The goal of this paper is to present how these metasurfaces can be used for industrial applications and where lies the challenges and opportunities that they bring.

Data-driven visual odometry (VO) is a critical subroutine for autonomous edge robotics, and recent progress in the field has produced highly accurate point predictions in complex environments. However, emerging autonomous edge robotics devices like insect-scale drones and surgical robots lack a computationally efficient framework to estimate VO's predictive uncertainties. Meanwhile, as edge robotics continue to proliferate into mission-critical application spaces, awareness of the model's predictive uncertainties has become crucial for risk-aware decision-making. This paper addresses this challenge by presenting a novel, lightweight, and statistically robust framework that leverages conformal inference (CI) to extract VO's uncertainty bands. Our approach represents the uncertainties using flexible, adaptable, and adjustable prediction intervals that, on average, guarantee the inclusion of the ground truth across all degrees of freedom (DOF) of pose estimation. We discuss the architectures of generative deep neural networks for estimating multivariate uncertainty bands along with point (mean) prediction. We also present techniques to improve the uncertainty estimation accuracy, such as leveraging Monte Carlo dropout (MC-dropout) for data augmentation. Finally, we propose a novel training loss function that combines interval scoring and calibration loss with traditional training metrics-mean-squared error and KL-divergence-to improve uncertainty-aware learning. Our simulation results demonstrate that the presented framework consistently captures true uncertainty in pose estimations across different datasets, estimation models, and applied noise types, indicating its wide applicability.

We present a two-step strategy to solve an inverse scattering problem in 2-D geometry. The first step approximates the inverse scattering as a convex optimization problem and provides an estimation of the total field inside the domain under investigation withouta prioriknowledge or tuning parameters. In the second step, the previously estimated total field is used to reconstruct the unknown contrast permittivity, which is represented by a superposition of level-1 Haar wavelet transform basis functions. Subject to${\ell _{1}}$-norm constraints of the wavelet coefficients, a least absolute shrinkage and selection operator (LASSO) problem that searches for the global minimum of the${\ell _{2}}$-norm residual is exploited by accounting for the sparsity of the wavelet-based permittivity representation. Numerical results are presented to assess the effectiveness of the proposed formulation against objects with relatively small electric size. Finally, the approach is validated against experimental data.

Pillars represent some of the commonest supporting elements of modern and historical buildings. Nondestructive testing methods can be applied to gain information about the status of these structural elements. Among them, ground penetrating radar (GPR) is a popular diagnostic tool for the assessment of concrete structures. Despite several theoretical and experimental studies on concrete structural evaluation by GPR have been reported, little work has been done so far with respect to pillars. Owing to their circular geometry, pillars are complex multiscattering environments, which render the interpretation of the radar images very challenging. This article deals with the application of radio frequency tomography as a nondestructive technique for imaging the inner structure of pillars. The main goal of the study is the assessment of the imaging performance that can be obtained in comparison to conventional GPR exploiting a multimonostatic configuration. Accordingly, potentialities and performance of multimonostatic and multiview/multistatic measurement configurations are herein investigated in the inverse scattering framework. For each measurement configuration, the regularized reconstruction of a point-like target and the spectral content are evaluated. The data inversion is carried out by means of the truncated singular value decomposition scheme. Tomographic reconstructions based on full-wave synthetic data are shown to support the comparative analysis.

Backscatter modulation (BM), usually used for radio frequency identifications, has the potential to be exploited in a wider range of applications such as the Internet of Things and wireless sensor networks. In this paper, we leverage the BM by increasing its down-link (from reader to tag) and up-link (from tag to reader) ranges. We propose two active two-way BM tag configurations, named parallel and series. For both configurations, we use active loads in the tag modulators to maximize the BM range and implement the desired backscattered constellation, subject to no data loss in the down-link path. Contrary to most existing BM studies, we use Thevenin/Norton equivalent circuit, only to calculate the received power at tag, while we derive the tag backscattered power using the antenna scatterer theorem. Moreover, we obtain a closed-form expression of the average bit error probability (BEP) at the reader in Rician fading channel environment for both tag configurations. We compare the proposed active BM tags with the conventional passive BM tags. The simulation results prove that for an average BEP equal to 10-4, an SNR improvement of up to 19 and 24 dB can be achieved for parallel and series configurations, respectively.

We present a new efficient control framework for controlling groups of heterogeneous stress-engineered MEMS microrobots for accomplishing micro-assembly. The objective is to maximize the number of controllable microrobots in the system while keeping the number of external global signals as low as possible. This work proposes a theoretical control strategy that could complete multiple-shapes microassembly from arbitrary initial configuration where all the control primitives can be accompanied with a constant number (O(1)) of control pulses of the power delivery waveform. We focus on microrobotic systems that can be modeled as nonholonomic unicycles. We validate the control policy with hardware experiments for implementing planar assembly using multiple macroscale robots with direct drive wheels. These results lay the foundation for developing new methods to control of a large number of MEMS microrobots.

In the near future, radar and communication systems will share the spectrum. This motivates the study of how the two systems, which have traditionally operated in different bands, may co-exist. This paper investigates the effect of radar interference (unaltered, beyond the communication system designer's control) on an uncoded communication system, using complex-valued modulation schemes when the Maximum-A-Posteriori (MAP) detector is used. For all commonly used higher order modulation schemes, the Symbol Error Rate (SER) exhibits an "error floor" for the radar interference much larger than the signal power, which can be exactly characterized; in this regime the optimal MAP detector behaves like an interference canceller; interestingly, in this regime the channel behaves as a real-valued phase- fading AWGN channel with receiver CSI, thus indicating a loss of one of the two complex dimensions compared to the complex-valued interference-free channel.

With the relentless increase in population density, the anthropogenic expansion into natural terrestrial hazard zones has become irreversible resulting in ever more catastrophic disasters, not only in the Asia-Pacific region more so within the entire tropical belt engulfing Mother Earth. Thus not only the Indonesian Pacific Islands, so also South America, Africa and back via the Islands of the Indian Ocean to Asia-Pacific, these natural events like volcano eruptions, earthquakes with emerging tsunami, cyclones and severe down pours, humidity and haze have caused havoc, loss of lives, destruction of infrastructure and above all intentional manmade interference resulting in the deterioration of pristine tropical jungle forests. Matters have become so bad that proposals are forthcoming for equating oil-palm and tropical-fruit orchard mono-cultures with pristine tropical jungle habitat by greedy developers mostly exterior to local environmental regions suffering helplessly from it.

Imaging of rebars inside a concrete block is investigated using the quadratic forward model of RF Tomography for three different antenna configurations. Reconstructed images are obtained using the method of moments.

An effective way to solve the inverse scattering from dielectric objects relies on the Born approximation, which allows to linearize the problem and retrieve a qualitative reconstruction of the targets in terms of location and extent. The limits of the validity of the linear model can be extended by considering a quadratic approximation of the operator relating the scattered field data to the unknown object function. The use of the quadratic operator allows on the one hand to recover additional spatial variations of the object profile and on the other hand to mitigate the local minima (false solution) problem typically affecting nonlinear inversion methods. In this letter, we present an experimental validation of the quadratic inverse model for dielectric objects in free space. The data processing confirms that the tomographic images based on the quadratic model are better resolved compared to the ones provided by the inversion of the linear Born model.

Software to compute angular and radial Mathieu functions is provided in the case that the parameterqis a complex variable and the independent variablexis real. After an introduction on the notation and the definitions of Mathieu functions and their related properties, Fortran 90 subroutines to compute them are described and validated with some comparisons. A sample application is also provided.

Radio-frequency (RF) tomography is extended for imaging underground structures and tunnels assuming rough terrain. The theory of RF tomography described in an earlier paper remains applicable, provided that a numerical Green's function is computed. An FFT-based and intrinsically parallel method for obtaining numerical Green's functions is described. This method is corroborated with explicit formulas and implemented for RF tomography. Simulated data computed using a finite-difference time-domain code are used to demonstrate performance.

Three extensions to radio-frequency (RF) tomography for imaging of voids under wide areas of regard are presented. These extensions are motivated by three challenges. One challenge is the lateral wave, which propagates in proximity of the air–earth interface and represents the predominant radiation mechanism for wide-area surveillance, sensing of denied terrain, or close-in sensing. A second challenge is the direct-path coupling between transmitters (Txs) and receivers (Rxs), that affects the measurements. A third challenge is the generation of clutter by the unknown distribution of anomalies embedded in the ground. These challenges are addressed and solved using the following strategies: 1) A forward model for RF tomography that accounts for lateral waves expressed in closed form (for fast computation); 2) a strategy that reduces the direct-path coupling between any Tx–Rx pair; and 3) an improved inversion scheme that is robust with respect to noise, clutter, and high attenuation. A finite-difference time domain simulation of a scenario representing close-in sensing of a denied area is performed, and reconstructed images obtained using the improved and the classical models of RF tomography are compared.

Radio frequency (RF) tomography is proposed to detect underground voids, such as tunnels or caches, over relatively wide areas of regard. The RF tomography approach requires a set of low-cost transmitters and receivers arbitrarily deployed on the surface of the ground or slightly buried. Using the principles of inverse scattering and diffraction tomography, a simplified theory for below-ground imaging is developed. In this paper, the principles and motivations in support of RF tomography are introduced. Furthermore, several inversion schemes based on arbitrarily deployed sensors are devised. Then, limitations to performance and system considerations are discussed. Finally, the effectiveness of RF tomography is demonstrated by presenting images reconstructed via the processing of synthetic data.

A translation to Fortran 90 of Gertrude Blanch's algorithm for computing the expansion coefficients of the series that represent Mathieu functions is presented. Its advantages are portability, higher precision, practicality of use, and extended documentation. In addition, numerical validations and comparisons with other existing methods are presented.