Search

Feedforward control with model inversion is a widely-used solution for high-precision output tracking. However, because inverting a non-minimum phase model generates unbounded control input, model-inversion only applies to limited types of systems. This paper presents a new non-parametric pseudo-inversion approach to design bounded optimal control input with desirable properties for arbitrary types of systems. Closed-form equations are presented for the batch (full preview) and recursive (limited preview) implementations of this approach, and its performance is compared against existing pseudo-inversion methods in benchmark numerical examples. Furthermore, the practical implementation of the proposed method is demonstrated by designing a feedforward controller for a commercial 3-Dimensional (3D) printer. The results show that the proposed approach effectively compensates for the structural vibrations of the printer, preventing layer-shifting errors that usually happen during high-speed printing. [ABSTRACT FROM AUTHOR]

Generative AI, including large language models (LLMs), has transformed the paradigm of data generation and creative content, but this progress raises critical privacy concerns, especially when models are trained on sensitive data. This review provides a comprehensive overview of privacy-preserving techniques aimed at safeguarding data privacy in generative AI, such as differential privacy (DP), federated learning (FL), homomorphic encryption (HE), and secure multi-party computation (SMPC). These techniques mitigate risks like model inversion, data leakage, and membership inference attacks, which are particularly relevant to LLMs. Additionally, the review explores emerging solutions, including privacy-enhancing technologies and post-quantum cryptography, as future directions for enhancing privacy in generative AI systems. Recognizing that achieving absolute privacy is mathematically impossible, the review emphasizes the necessity of aligning technical safeguards with legal and regulatory frameworks to ensure compliance with data protection laws. By discussing the ethical and legal implications of privacy risks in generative AI, the review underscores the need for a balanced approach that considers performance, scalability, and privacy preservation. The findings highlight the need for ongoing research and innovation to develop privacy-preserving techniques that keep pace with the scaling of generative AI, especially in large language models, while adhering to regulatory and ethical standards. [ABSTRACT FROM AUTHOR]

Nitrogen is one of the most important macronutrients for plant growth and timely estimation of canopy nitrogen content (CNC) is crucial for agricultural applications. Remote sensing has emerged as an important tool to quantify CNC using either empirically or physically based methods. Most empirical methods use chlorophyll related spectral indices and are dependent on the relationship between nitrogen and chlorophyll, which varies with vegetation types and growth stages. In contrast, physically based methods use the full-range reflectance data and retrieve CNC from coupled leaf and canopy radiative transfer models (such as PROSPECT-PRO + 4SAIL, PROSAIL-PRO). However, the subtle absorption features of nitrogen and protein in fresh leaves hinder the accurate estimation of CNC. Therefore, this study proposed an efficient and mechanistic framework to estimate CNC (PROSAIL-NAM) by coupling PROSAIL-PRO with a nitrogen allocation model, which divided the total nitrogen into non-photosynthetic nitrogen (NPN) and photosynthetic nitrogen (PN). At the canopy level, PN and NPN are assumed to be proportional to canopy chlorophyll content (CCC) and canopy dry matter content (CDM), respectively. The PROSAIL-PRO model was first used to estimate CCC and CDM, and then the resulting CCC and CDM were fed to the nitrogen allocation model to estimate CNC. The estimation accuracy of CNC was assessed with six diverse datasets: four from field crop experiments across geographic sites, one from multiple ecosystems, and one from a satellite-ground joint experiment. Our results showed that satisfactory estimations of CNC were obtained when CCC and CDM were estimated using a model inversion method (RMSE = 0.54–1.56 g/m2) and a hybrid retrieval method (RMSE = 0.49–2.25 g/m2). The model inversion method was comparable with the hybrid retrieval method for ground platforms, but performed better for airborne and satellite platforms. In addition, the traditional protein-nitrogen conversion model obtained CNC from the canopy protein content and led to clear overestimations of CNC with RMSE > 1.95 g/m2. This study represents a first attempt to develop a robust approach by coupling PROSAIL-PRO with a nitrogen allocation model for accurate estimation of CNC across geographic sites, ecosystems, and platforms. These finding will advance the monitoring of CNC from regional to global scales.

Over the past six decades, carbonyl sulfide (COS) in terrestrial ecosystems has been extensively studied, with research focusing on exploring its ecological and environmental effects, estimating source–sink volume, and identifying influencing factors. The global terrestrial COS sink has been estimated to be about 1.194–1.721 Tg a−1, with the terrestrial sink induced by plants and soils 0.50–1.20 Tg a−1, accounting for 41%–69% of the total. Hence, the role of plants and soils as COS sinks has been extensively explored. Now we know that factors such as the activity of carbonic anhydrase (CA), leaf structural traits, soil microbial activity, and environmental factors play significant roles in the COS budget. Developments in observational techniques have also made important contributions to the COS budget. This paper provides an overview of the research progress made on COS based on a comprehensive review of the literature. Then, it highlights the current research hotspots and issues requiring further exploration. For instance, it has been demonstrated that there are still significant uncertainties in the estimation of COS sources and sinks, emphasizing the need for further exploration of COS measuring techniques. This review aims to provide comprehensive guidance for COS research in terrestrial ecosystems. [ABSTRACT FROM AUTHOR]

Today, the significance of biometrics is more pronounced than ever in accurately allowing access to valuable resources, from personal devices to highly sensitive buildings, as well as classified information. Researchers are pushing forward toward devising robust biometric systems with higher accuracy, fewer false positives and false negatives, and better performance. On the other hand, machine learning (ML) has been shown to play a key role in improving such systems. By constantly learning and adapting to users' changing biometric patterns, ML algorithms can improve accuracy and performance over time. The integration of ML algorithms with biometrics, however, introduces vulnerabilities in such systems. This article investigates the new issues of concern that come about because of the adoption of ML methods in biometric systems. Specifically, techniques to breach biometric systems, namely, data poisoning, model inversion, bias injection, and deepfakes, are discussed. Here, the methodology consisted of conducting a detailed review of the literature in which ML techniques have been adopted in biometrics. In this study, we included all works that have successfully applied ML and reported favorable results after this adoption. These articles not only reported improved numerical results but also provided sound technical justification for this improvement. There were many isolated, unsupported, and unjustified works about the major advantages of ML techniques in improving security, which were excluded from this review. Though briefly mentioned, we did not touch upon encryption/decryption aspects, and, accordingly, cybersecurity was excluded from this study. At the end, recommendations are made to build stronger and more secure systems that benefit from ML adoption while closing the door to adversarial attacks. [ABSTRACT FROM AUTHOR]

Handling suspended loads in cluttered environments is critical due to the oscillations arising while the load is traveling. Exploiting active control algorithms is often unviable in industrial applications, due to the necessity of installing sensors on the load side, which is expensive and often impractical due to technological limitations. In this light, this paper proposes a trajectory planning method for underactuated, non-flat, non-minimum phase spatial gantry crane moving in structured cluttered environments. The method relies on model inversion. First, the system dynamics is partitioned into actuated and unactuated coordinates and then the load displacements are described as a non-linear separable function of these. The unactuated dynamic is unstable; hence, the displacement, velocity, and acceleration references are modified through the output redefinition technique. Finally, platform trajectory is computed, and the desired displacements of the load are obtained. The effectiveness of the proposed method is assessed through numerical and experimental tests performed on a laboratory testbed composed by an Adept Quattro robot moving a pendulum. The load is moved in a cluttered environment, and collisions are avoided while simultaneously tracking the prescribed trajectory effectively. [ABSTRACT FROM AUTHOR]

Purpose: This paper proposes a motion planning technique for precise path and trajectory tracking in an underactuated, non-minimum phase, spatial overhead crane. Besides having a number of independent actuators that is smaller than the number of degrees of freedom, tip control on this system presents unstable internal dynamics that leads to divergent solution of the inverse dynamic problem. Method: The paper exploits the representation of the controlled output as a separable function of the actuated (i.e., the platform translations) and unactuated (i.e., the swing angles) coordinates to easily formulate the internal dynamics, without any approximation, and to study its stability. Then, output redefinition is adopted within the internal dynamics to stabilize it, leading to stable and causal reference commands for the platform translations. Results: Besides proposing the theoretical formulation of this novel method, the paper includes the numerical validation and the experimental application on a laboratory setup. Comparison with the state-of-the-art input shaping is also proposed. Conclusion: The results, obtained through different reference trajectories, clearly show that almost exact tracking is obtained also in the experiments, by outperforming the benchmarks. [ABSTRACT FROM AUTHOR]

Handling suspended loads in cluttered environments is critical due to the oscillations arising while the load is traveling. Exploiting active control algorithms is often unviable in industrial applications, due to the necessity of installing sensors on the load side, which is expensive and often impractical due to technological limitations. In this light, this paper proposes a trajectory planning method for underactuated, non-flat, non-minimum phase spatial gantry crane moving in structured cluttered environments. The method relies on model inversion. First, the system dynamics is partitioned into actuated and unactuated coordinates and then the load displacements are described as a non-linear separable function of these. The unactuated dynamic is unstable; hence, the displacement, velocity, and acceleration references are modified through the output redefinition technique. Finally, platform trajectory is computed, and the desired displacements of the load are obtained. The effectiveness of the proposed method is assessed through numerical and experimental tests performed on a laboratory testbed composed by an Adept Quattro robot moving a pendulum. The load is moved in a cluttered environment, and collisions are avoided while simultaneously tracking the prescribed trajectory effectively.

Collaborative machine learning settings such as federated learning can be susceptible to adversarial interference and attacks. One class of such attacks is termed model inversion attacks, characterised by the adversary reverse-engineering the model into disclosing the training data. Previous implementations of this attack typically only rely on the shared data representations, ignoring the adversarial priors, or require that specific layers are present in the target model, reducing the potential attack surface. In this work, we propose a novel context-agnostic model inversion framework that builds on the foundations of gradient-based inversion attacks, but additionally exploits the features and the style of the data controlled by an in-the-network adversary. Our technique outperforms existing gradient-based approaches both qualitatively and quantitatively across all training settings, showing particular effectiveness against the collaborative medical imaging tasks. Finally, we demonstrate that our method achieves significant success on two downstream tasks: sensitive feature inference and facial recognition spoofing. [ABSTRACT FROM AUTHOR]

Karst wetlands are widely distributed throughout the southwest China, and play an important role in enhancing carbon sequestration and improving water quality in karst areas. The internationally important karst wetland of Huixian is the largest karst wetland in China, but its water quality has continued to deteriorate as a result of human influences in recent years. Remote sensing technology has become an important approach to estimate water quality parameters (WQPs). However, the feasibility of combining multi-sensor remote sensing images with deep learning to estimate different WQPs in karst wetlands has not been demonstrated yet. To resolve this issue, this study constructed multiple retrieval models of WQPs (Chlorophyll-a (Chla), Phycocyanin (PC), Turbidity (Turb), Dissolved Oxygen (DO)) in karst wetlands using deep learning (Transformer and Mixture Density Network (MDN)) and optimized shallow machine learning (Random Forest (RF), XGBoost (XGB) and Gradient Boosting (GB)) based on multi-sensor images from satellite and UAV platforms. The performance of deep learning in the inversion of WQPs demonstrated to compare with shallow machine learning using multispectral and hyperspectral images. We further quantitatively evaluated the retrieval performance of UAV and satellite, multispectral and hyperspectral images, and presented predictive mapping of the gradient distribution of WQPs. Finally, this study adopted the SHapley Additive exPlanations (SHAP) to tackle the local and global interpretability of the input features contribution to the output of retrieval models. The results showed that (1) Transformer model presented a good prediction of PC and DO (R2 = 0.649 ∼ 0.844), XGB and GB models achieved the highest accuracy estimation of Chla and Turb (R2 = 0.75). (2) The estimation results of WQPs based on UAV platform (R2 = 0.419 ∼ 0.695) was higher than that of satellite-based images. The estimation accuracy of multispectral images (R2 = 0.338 ∼ 0.718) was slightly higher than that of Zhuhai-1 Orbita hyperspectral (OHS) images. The average accuracy of Turb estimated by UAV images (R2 = 0.565 ∼ 0.752) was higher than that of satellite-based images. OHS hyperspectral images had the best DO estimation (R2 = 0.314 ∼ 0.649). (3) This study found 32.66% and 23.01% of water area with the Chla and Turb concentrations exceeding 60 μg/L and 60 NTU, respectively, which revealed that the Huixian karst wetland has suffered serious water pollution. (4) The SHAP analysis reveals that near infra-red and red band are sensitive to predict Chla and DO, red and red-edge bands are sensitive to predict PC and Turb in the karst wetland.

Soil moisture estimation from agriculture fields using SAR measurements is a challenging process owing to the presence of vegetation canopy. In this study, the soil moisture (SM) is retrieved from multi-polarization airborne L- and C-band E-SAR data of different agriculture fields by using the radar parameter, Radar Vegetation Index (RVI). The retrieval methodology employs the semi-empirical Water Cloud Model (WCM) for vegetation-soil system modeling, followed by an inversion algorithm based on a Look Up Table approach. The impact of using different vegetation descriptors, both from in situ measured (Leaf Area Index, Wet Biomass and Vegetation Water Content) and radar derived (L-band RVI and C-band RVI), on the WCM inversion for SM retrieval is examined. The use of the RVI as the vegetation descriptor, which is obtained from C-band data, improves soil moisture retrieval with an RMSE of 7–8% volumetric soil moisture at L-band. [ABSTRACT FROM AUTHOR]

This paper deals with Feedback/Feedforward (FB/FF) control of a gantry crane system intended for the transport of payloads that take values over a known interval. It is also assumed that the crane is affected by unmeasurable disturbances. A new 2DoF control architecture is proposed whose purpose is to speed up the horizontal payload transition while minimizing its oscillations. The main features of the control design procedure are as follows: (1) The output FB controller is designed to ensure robust closed loop stability and steady‐state exact payload positioning; (2) the disturbance is estimated by means of an observer, and its transient effect is compensated through the FF action; and (3) the robust FF control action is given by the optimally weighted sum of the two contributions due to FF Plant Inversion (FFPI) and FF Closed Loop Inversion (FFCLI) control schemes. [ABSTRACT FROM AUTHOR]

The exploitation of mineral resources has seriously polluted the environment around mines, notably in terms of heavy metal contamination of tailings pond soil. Hyperspectral remote sensing, as opposed to conventional on-site sampling and laboratory analysis, offers a potent tool for effective monitoring the content of soil heavy metals. Therefore, we investigated the inversion models of heavy metal content in metal tailings area based on measured hyperspectral and multispectral data. Hyperspectral and its transformation, as well as the simulated Landsat8-OLI multispectral were used for model inversion respectively. Stepwise Multiple Linear Regression (SMLR), Partial Least Squares Regression (PLSR) and Back Propagation Neuron Network (BPNN) were established to study the spectral inversion of eight heavy metals (Cu, Cd, Cr, Ni, Pb, Zn, As, and Hg). The direct inversion models were established on the basis of correlation analysis and the adjust coefficient of determination (Adjust_R2) and Root Mean Square Error (RMSE) were used for model evaluation. Then the best combination of spectral transformation and inversion model were explored. The model inversion results suggested that: (1) Hyperspectral transformation can generally improve the model accuracy, especially the second derivative spectral, based on which the training Adjust_R2 of Hg SMLR and PLSR models are as high as 0.795 and 0.802. (2) The BP neural network inversion based on the denoised hyperspectrum demonstrate that both the training and testing Adjust_R2 of Cd, Ni and Hg models are all greater than 0.5, indicating good applicability in practical extrapolation. (3) Both the training and testing Adjust_R2 of Cu and Hg PLSR models based on simulated R_Landsat8-OLI multispectral are greater than 0.5, and Hg has lower RMSE and lager Adjust_R2 with training and testing Adjust_R2 values of 0.833 and 0.553 respectively. (4) Multispectral remote sensing detection and mapping of Hg contamination were realized by the optimal simulation model of Hg. Hence, it is feasible to simulate the multispectral with hyperspectral data for investigating heavy metal contamination.

Through the constant rain of sinking marine particles in the ocean, carbon (C) trapped within is exported into the water column and sequestered when reaching depths below the mesopelagic zone. Atmospheric CO2 levels are thereby strongly related to the magnitude of carbon export fluxes in the mesopelagic zone. Sinking particles represent the main source of carbon and energy for mesopelagic organisms, attenuating the C export flux along the water column. Attempts to quantify the amount of C exported versus consumed by heterotrophic organisms have increased in recent decades. Yet, most of the conducted estimations have led to estimated C demands several times higher than the measured C export fluxes. The choice of parameters such as growth efficiencies or various conversion factors is known to greatly impact the resulting C budget. In parallel, field or experimental data are sorely lacking to obtain accurate values of these crucial overlooked parameters. In this study, we identify the most influential of these parameters and perform inversion of a mechanistic model. Further, we determine the optimal parameter values as the ones that best explain the observed prokaryotic respiration, the prokaryotic production, and the zooplankton respirations. The consistency of the resulting C-budget suggests that such budgets can be adequately balanced when using appropriate parameters. [ABSTRACT FROM AUTHOR]

Stable inversion represents a classic approach to achieve the exact inverse input for non-minimum-phase (NMP) systems. Solutions deduced based on state–space equations and transfer functions have been frequently proposed, however they are basically developed in the infinite-time horizon and are inherently time-domain computation methods. Considering the practical finite-time tracking tasks, this paper investigates the finite-time stable inversion problem. In particular, the discrete-frequency-domain solution which enables frequency-domain computation is studied. As the main contribution of the paper, the accuracy issues of the discrete-frequency-domain solution are revealed and an easy-to-use procedure is provided to improve the inversion accuracy by utilizing pre-actuation and post-actuation methods. Simulation and experiment both verify the effectiveness of the developed discrete-frequency-domain stable inversion technique. • The finite-time stable inversion problem is investigated for the NMP systems. • The discrete-frequency-domain solution suffers from the initial-state issue and the leakage-error issue. • The pre-actuation and post-actuation techniques can improve the accuracy of the discrete-frequency-domain solution. [ABSTRACT FROM AUTHOR]

Recently, a popular line of research in face recognition is adopting margins in the well-established softmax loss function to maximize class separability. In this paper, we first introduce an Additive Angular Margin Loss (ArcFace), which not only has a clear geometric interpretation but also significantly enhances the discriminative power. Since ArcFace is susceptible to the massive label noise, we further propose sub-center ArcFace, in which each class contains $K$ K sub-centers and training samples only need to be close to any of the $K$ K positive sub-centers. Sub-center ArcFace encourages one dominant sub-class that contains the majority of clean faces and non-dominant sub-classes that include hard or noisy faces. Based on this self-propelled isolation, we boost the performance through automatically purifying raw web faces under massive real-world noise. Besides discriminative feature embedding, we also explore the inverse problem, mapping feature vectors to face images. Without training any additional generator or discriminator, the pre-trained ArcFace model can generate identity-preserved face images for both subjects inside and outside the training data only by using the network gradient and Batch Normalization (BN) priors. Extensive experiments demonstrate that ArcFace can enhance the discriminative feature embedding as well as strengthen the generative face synthesis. [ABSTRACT FROM AUTHOR]

We apply the concept of servo constraints to end-effector trajectory tracking control of parallel robots with structural link flexibilities. Such servo constraints deliver the inverse robot model where solution approaches via projections are proposed, which transform the resulting differential-algebraic equations to ordinary differential equations. The applicable solution process depends on the existence and stability of the internal dynamics. When using the exact end-effector of flexible link robots as output, this internal dynamics is usually unstable. Then a two-point boundary value problem is considered in the framework of stable inversion to obtain the noncausal solution offline. This solution is used as a feedforward control, which is initially combined only with actuator feedback control. To also account for errors within the link flexibility, the well-known linear–quadratic regulator is adapted to end-effector trajectory tracking based on differential-algebraic equations. Finally, we propose a systematic input–output feedback linearization approach, which uses servo constraints for flexible link parallel robots. Here a minimum phase system is obtained by tracking a redefined end-effector output, which is an approximation of the exact end-effector position. All control concepts are validated experimentally with a parallel robot having a highly flexible link. The results allow us to compare different control approaches and show the superior performance of controllers that rely on a flexible multibody model in contrast to classical rigid multibody modeling. [ABSTRACT FROM AUTHOR]

Federated learning (FL) is gradually becoming a key learning paradigm in Privacy‐preserving Machine Learning (ML) systems. In FL, a large number of clients cooperate with a central server to learn a shared model without sharing their own data sets. However, since there is a great disparity between the client data sets, standard FL is often hard to tune and suffers from performance degradation due to the inharmony among local models. To this end, in this paper we propose a novel FL scheme, termed client reputation federated learning (CRFL), which dynamically assesses the reputation of the clients participating in FL. Our method leverages techniques from model explanation, and aims at precisely measure each client's impact to the global model. To be specific, we first calculate the saliency‐weighted variance on pixelwise relevance scores as the quality factor of a single sample. Then we extract activation function values at the last hidden layer to compute the divergence factor of individual data set. Finally, the server integrates these two factors as an assessment of the client reputation. By leveraging such assessment, CRFL can dynamically adjust the weights of the clients in each aggregation round, thus leading to a significant improvement over the baseline method in terms of model accuracy and convergence rate. Intensive experiments are conducted on the MNIST and CIFAR‐10 data sets, and experimental results demonstrate the efficacy of the proposed method. [ABSTRACT FROM AUTHOR]

In this era of global warming, the regular and accurate mapping of vegetation conditions is essential for monitoring ecosystems, climate sustainability and biodiversity. In this context, this work proposes a physics-guided data-driven methodology to invert radiative transfer models (RTM) for the retrieval of vegetation biophysical variables. A hybrid paradigm is proposed by incorporating the physical model to be inverted into the design of a neural network architecture, which is trained by exploiting unlabeled satellite images. In this study, we show how the proposed strategy allows the simultaneous probabilistic inversion of all input PROSAIL model parameters by exploiting Sentinel-2 images. The interest of the proposed self-supervised learning strategy is corroborated by showing the limitations of existing simulation-trained machine learning algorithms. Results are assessed on leaf area index (LAI) and canopy chlorophyll content (CCC) in-situ measurements collected on four different field campaigns over three European tests sites. Prediction accuracies are compared with performances reached by the well-established Biophysical Processor (BP) of the Sentinel Application Platform (SNAP). Obtained overall accuracies corroborate that the proposed methodology achieves performances equivalent to or better than the state-of-the-art methods. • Full Bayesian inversion of PROSAIL based on Variational Autoencoders. • Self-supervised learning on Sentinel-2 images. • Experimental correlations found between PROSAIL variables. • Existing approaches highly dependent on the training data simulations distributions. • In-situ LAI and CCC retrieval equivalent or better than SNAP's SL2P. [ABSTRACT FROM AUTHOR]

Understanding plant hydraulic regulation is critical for predicting plant and ecosystem responses to projected increases in drought stress. Plant hydraulic regulation is controlled by observable, diverse plant hydraulic traits that can vary as much across individuals of the same species as they do across different species. Direct measurements of plant hydraulic traits from a range of ecosystems remain limited in comparison to other, more readily measured traits (e.g., specific leaf area). Furthermore, plant hydraulic trait measurements, often made at leaf or branch levels, are not easily scaled to whole‐plant values that are typically used to predict plant and ecosystem fluxes. In this study, multiple whole‐plant hydraulic parameters are inferred from observations of plant water use (i.e., sap flow), soil properties, and meteorological data. We use a Markov Chain Monte Carlo model inversion approach to obtain the best estimates and uncertainty of plant hydraulic parameters that capture whole‐plant effective embolism resistance and stomatal sensitivity to decreasing plant water potential. We then use the inferred values in the model to estimate whole‐tree water use and isohydricity. This approach reliably infers whole‐plant parameter values with enough specificity to resolve inter‐ and intra‐specific differences, and thus supplements time‐ and labor‐intensive direct measurements of traits. Key Points: Plant hydraulic parameters are inferred with low uncertainty from sap flow dataInferred parameter values capture whole‐plant response and water use strategies instead of leaf or branch‐level responsesThe model inversion method complements field measurement of plant hydraulic traits [ABSTRACT FROM AUTHOR]

This article deals with the issue of tracking a reference optimal trajectory for an autonomous nonlinear vehicle model by designing both lateral and longitudinal robust feedback control and a suited feedforward control. In previous works, a strategy based on a human-driver field of view was used to plan an optimal trajectory reference. The optimization has been made using a Genetic Algorithm (GA), and the obtained trajectory has been injected into a Potential Field (PF) so as to be reactive to unforeseen events by using a point mass model. Here, the previously developed GA-PF planification process is integrated in a new complete global planning and tracking method and applied to a validation nonlinear vehicle model. This control tracking method is developed in two strategies: a bicycle model is used as model inversion for feedforward design and a robust control is designed as feedback control in order to take the vehicle (mass and velocity) and road (slope and adherence) parameter variations into account. A lateral nonlinear control and a longitudinal robust control are designed. Realistic autonomous car simulation results are provided on an overtaking scenario and a round-about scenario. [ABSTRACT FROM AUTHOR]

This paper deals with FeedBack/ FeedForward (FB/FF) control of double pendulum gantry crane systems with payloads taking values over arbitrarily large intervals. The new proposed 2DoF control architecture is aimed at: 1) to speed up the horizontal payload transportation while minimizing the tracking error with respect to a desired trajectory; 2) to minimize the sway angles amplitude. The main features of the control design procedure are: 1) the dynamic output FB control is designed in order to ensure the robust stability of the closed loop system and the steady-state exact payload positioning; 2) the FF control action is given by the optimally weighted sum of the two contributions due to FF Plant Inversion (FFPI) and FF Closed Loop Inversion (FFCLI) control schemes; 3) the optimal robust FF control input is obtained as the solution of a min max optimization problem that can be solved offline with numerically efficient procedures; 4) the provided analytical closed form of the FF input in terms of a linear combination of polynomial B-splines basis functions allows an easy implementation on commercial devices.

Rainfall-runoff modeling is highly uncertain for a number of different reasons. Hydrological processes are quite complex, and their simplifications in the models lead to inaccuracies. Model parameters themselves are uncertain—physical parameters because of their observations and conceptual parameters due to their limited identifiability. Furthermore, the main model input—precipitation is uncertain due to the limited number of available observations and the high spatio-temporal variability. The quantification of model output uncertainty is essential for their use. Most approaches used for the quantification of uncertainty in rainfall-runoff modeling assign the uncertainty to the model parameters. In this contribution, the role of precipitation uncertainty is investigated. Instead of a standard sensitivity analysis of the model output with respect to the input variations, it is investigated to what extent realistic precipitation fields could improve model performance. Realistic precipitation fields are defined as gridded realizations of precipitation which reproduce the observed values at the observation locations, with values which reproduce the distribution of the observed values and with spatial variability the same as the spatial variability of the observations. The above conditions apply to each observation time step. Through an inverse modeling approach based on Random Mixing precipitation fields fulfilling the above conditions and reproducing the discharge output better than using traditional interpolated observations can be obtained. These realizations show how much rainfall runoff models may profit from better precipitation input and how much remains for the parameter and model concept uncertainty. The methodology is applied using two hydrological models with a contrasting basis, SHETRAN and HBV, for three different mesoscale sub-catchments of the Neckar basin in Germany. Results show that up to 50% of the model error can be attributed to precipitation uncertainty. Further, inverting precipitation using hydrological models can improve model performance even in neighboring catchments which are not considered explicitly.

This paper focuses on the problem of determining the most appropriate Two Degrees of Freedom (2DoF) control architecture, when the FeedForward (FF) action is the result of a stable model inversion procedure. The purpose is to define a control scheme with enhanced tracking performance even in the case of non minimum phase MIMO plant affected by polytopic uncertainty and with a possible non hyperbolic internal dynamics. The new proposed 2DoF architecture is given by an optimal balance of the control actions produced by FeedForward Plant Inversion (FFPI) and FeedForward Closed Loop Inversion (FFCLI). This new architecture is referred to as FeedForward Optimally Balanced Inversion (FFOBI). Robustness with respect to polytopic uncertainty is obtained using a min-max optimization approach. Numerical results show that the FFOBI improves the tracking of both FFPI and FFCLI. [ABSTRACT FROM AUTHOR]

We propose two methods to calibrate the parameters of the epidemic‐type aftershock sequence (ETAS) model based on expectation maximization (EM) while accounting for temporal variation of catalog completeness. The first method allows for model calibration on long‐term earthquake catalogs with temporal variation of the completeness magnitude, mc. This calibration technique is beneficial for long‐term probabilistic seismic hazard assessment (PSHA), which is often based on a mixture of instrumental and historical catalogs. The second method generalizes the concept of mc, considering rate‐ and magnitude‐dependent detection probability, and allows for self‐consistent estimation of ETAS parameters and high‐frequency detection incompleteness. With this approach, we aim to address the potential biases in parameter calibration due to short‐term aftershock incompleteness, embracing incompleteness instead of avoiding it. Using synthetic tests, we show that both methods can accurately invert the parameters of simulated catalogs. We then use them to estimate ETAS parameters for California using the earthquake catalog since 1932. To explore how model calibration, inclusion of small events, and accounting for short‐term incompleteness affect earthquakes' predictability, we systematically compare variants of ETAS models based on the second approach in pseudo‐prospective forecasting experiments for California. Our proposed model significantly outperforms the ETAS null model, with decreasing information gain for increasing target magnitude threshold. We find that the ability to include small earthquakes for simulation of future scenarios is the primary driver of the improvement and that accounting for incompleteness is necessary. Our results have significant implications for our understanding of earthquake interaction mechanisms and the future of seismicity forecasting. Plain Language Summary: Our capability to detect earthquakes varies with time, on one hand because more and better instruments are being deployed over time, leading to long‐term changes of detection capability. On the other hand, earthquakes are more difficult to be detected when seismic activity is high, which manifests in short‐term changes of detection capability. Incomplete detection can lead to biases in epidemic‐type aftershock sequence (ETAS) models used for earthquake forecasting. We propose two methods that allow us to calibrate these models while accounting for long‐term (first method) and short‐term (second method) changes in detection capability, which allows us to use a larger and more representative fraction of the available data. We test both methods on synthetic data and then apply them to the Californian earthquake catalog. Using the second method, we test how small earthquakes can improve our forecasts. We find that the ability to include small earthquakes in simulations leads to superior forecasts, and that it is necessary to correct for short‐term incompleteness to achieve this superiority. The positive effect is strongest when forecasting relatively small events, and decreases when forecasting larger events. These results have important implications for our understanding of earthquake interactions and for the future of earthquake forecasting. Key Points: Two methods are proposed to invert ETAS parameters when catalog completeness varies with timeUsing pseudo‐prospective experiments we compare the forecasting skill of our proposed models to a strong ETAS null modelIncluding small events in simulations yields increasingly superior forecasts for decreasing target magnitudes when using STAI correction [ABSTRACT FROM AUTHOR]

Hyperspectral lidar (HSL) is a novel remote sensing technology that provides spectral information in addition to spatial features. This unprecedented data source leads to new possibilities for monitoring leaf biochemistry. Inversion of physically based radiative transfer models (RTMs) is a popular method for deriving leaf physiological traits due to its robustness and generalization capability. However, owing to the active nature of the HSL system, RTM inversion using the backscattered reflectance spectra may face new problems. Thus, optimization strategies for RTM inversion based on HSL measurements need to be studied. In this paper, several regulation strategies for lookup table (LUT)-based PROSPECT model inversions were explored for an HSL system. In particular, the influences of i) different cost functions, ii) multiple best solutions (1–1000), iii) different LUT sizes (100–100000), and iv) spectral domains for leaf chlorophyll (Chl) retrieval were analyzed. An evaluation against an experimental dataset of rice leaves indicated that i) least-squares estimation (LSE) provided better estimates than seven alternative cost functions when more than 200 solutions were taken; ii) accuracy in leaf Chl retrieval increased up until 200 solutions where after it stabilized; iii) the impact of LUT size became insignificant after 1000; and iv) the red edge was the spectral domain that had the largest impact on the inversion performance. The optimal performance of leaf Chl estimation reached R2 of 0.58 and RMSE of 0.69 between the z-scores from retrieved and measured leaf Chl. The practical application of combining RTM with HSL data will facilitate the detection of leaf-level biochemistry and advance research on terrestrial carbon cycle modeling.

Fuel moisture content (FMC) of live vegetation is a crucial wildfire risk and spread rate driver. This study presents the first daily FMC product at a global scale and 500 m pixel resolution from the Moderate Resolution Imaging Spectroradiometer (MODIS) and radiative transfer models (RTMs) inversion techniques. Firstly, multi-source information parameterized the PROSPECT-5 (leaf level), 4SAIL (grass and shrub canopy level) and GeoSail (tree canopy level) RTMs to generate three look-up tables (LUTs). Each LUT contained the most realistic model inputs range and combination, and the corresponding simulated spectra. Secondly, for each date and location of interest, a global landcover map classified fuels into three classes: grassland, shrubland and forest. For each fuel class, the best LUT-based inversion strategy based on spectral information, cost function, percentage of solutions, and central tendency determined the optimal model for the global FMC product. Finally, 3,034 FMC measurements from 120 worldwide sites validated the statistically significant results (R2 = 0.62, RMSE = 34.57%, p

Quantitative determination of the physical properties of natural granite has been attempted from remotely sensed information, for which the Hapke model is a popular method. However, using the model to retrieve the photometric properties of terrestrial rocks (slab or particulate samples), especially for those with complex surface conditions such as natural granite, remains a challenge. In this study, we have approached the dilemma by coupling both radiative transfer (Hapke’s isotropic multiple scattering approximation (IMSA) model) and an empirical relationship between particle sizes with its critical parameter, the single-scattering albedo (SSA, $\omega$), determined from bidirectional reflectance (BDR) measurements. The results clearly indicated that the particle size of natural granite systematically controlled the BDR, which can be well fit by the Hapke model with varying parameters. The retrieved photometric parameters of the coefficients in the phase function ($b$ and $c$) can effectively indicate the scattering behavior of natural granite, but the variations in their values did not strongly correlate with the change in particle sizes. Instead, a good linear relationship between the SSA values and particle sizes has been established. By coupling the relationship into the Hapke model, we found a practical approach to estimate the particle size for measured samples from inversely retrieved SSA. Through this method, we are able to retrieve the physical properties of granite under natural surface conditions, and we foresee that the approach will be widely used in the future. [ABSTRACT FROM AUTHOR]

Long Short-Term Memory (LSTM) neural networks are well suited for representing time series as, compared to other neural networks, their structure avoids vanishing or exploding gradients. LSTM has been embedded into Model Predictive Control algorithms in order to forecast the behavior of nonlinear systems. The new algorithm presented in the paper is of a different nature, as the LSTM network approximates the inverse of the system over a receding horizon and provides a sequence of future inputs as a function of a specified output trajectory. The main advantage of the method appears when the desired output trajectory is generated from a small set of parameters, for example, a convergence rate. The Model Predictive control optimizes its criterion with respect to this small set of variables, and the LSTM supplies the corresponding future control inputs. Eventually, the modeling error of the LSTM can be compensated by feeding the control sequence to the forward model and updating the controller according to the output deviation. The algorithm allows to design Model Predictive controllers for nonlinear systems in a generic way, using a very small number of decision variables even with a long receding horizon. [ABSTRACT FROM AUTHOR]

Forest canopies exhibit spatial heterogeneity that impacts the relationship between essential climate variables such as leaf area index (LAI) or the fraction of absorbed photosynthetically active radiation (fAPAR) and bi-directional surface reflectance, and subsequently the estimation of these variables from satellite measurements. The Simplified Level 2 Prototype Processor (SL2P) allows global LAI and fAPAR mapping at 20 m resolution using Sentinel 2 imagery. Previous validation studies over forests indicate SL2P underestimates LAI by up to 50% in comparison to in-situ reference measurements. Our study tests the hypothesis that the SL2P LAI and fAPAR bias can be reduced by replacing the spatially homogenous SAILH canopy radiative transfer model used to calibrate SL2P with the heterogenous 4SAIL2 model, together with a shoot clumping parameterization. We also hypothesized that the additional parameters involved in this new version of SL2P (SL2P-CCRS) would lead to an increase in precision error and subsequently a bias-variance trade off. SL2P-CCRS reduced LAI bias by 65%, in comparison to SL2P, during direct validation with 1107 in-situ measurements. The LAI absolute bias reduced by ∼0.5 at LAI 3 and by ∼1 at LAI 6. SL2P-CCRS reduced fAPAR bias by 31% compared to SL2P but <0.05 on an absolute basis. Bias reduction was accompanied by an increase in precision error so that overall uncertainty, quantified by the root mean square difference in comparison to in-situ measurements, reduced by only 6% for LAI and 5% for fAPAR. These findings support the hypothesis that updating SL2P with a spatially heterogeneous RTM can reduce LAI and fAPAR bias over forests. The results also support the hypothesis that there is a bias-variance trade-off for LAI, and to a lesser extent for fAPAR, when increasing the complexity of SL2P by using a radiative transfer model that accounts for spatial heterogeneity. Nevertheless, SL2P-CCRS increased the agreement rate with Global Climate Observing System uncertainty requirements from 52% to 58% for LAI and 32% to 40% for fAPAR, suggesting that the trade-off is worthwhile, and that algorithms such as SL2P-CCRS, that use a spatially heterogenous radiative transfer model, should be applied for mapping fAPAR and LAI from Sentinel-2 measurements. • Updated SL2P algorithm with heterogeneous radiative transfer model (SL2P-CCRS). • SL2P-CCRS reduced forest LAI (fAPAR) bias by 65% (31%) compared to SL2P. • Total uncertainty reduced by only ∼5% due to an increase in precision error. • SL2P-CCRS meets GCOS LAI and fAPAR requirements over forests for LAI < 4. [ABSTRACT FROM AUTHOR]

N-linked glycan distribution plays a significant role in the generation of therapeutic proteins. It is challenging to determine the operating conditions when developing a new biopharmaceutical product with the desired glycan distributions. The glycosylation is a high complex nonlinear system, and it is difficult to develop a reliable first-principle model that heavily relies on experimentation. Our goal is to develop a nonlinear data-driven model and find an appropriate operating space included kinds of input combination from process variables based on this model to ensure the desired product quality. A methodology is proposed based on the inversion of a nonlinear latent-variable model (locality preserving projection to latent structures, LPPLS) to identify a subspace of the knowledge space. The normal operating points of the input variables are designed based on the LPPLS inversion, and the range of operating conditions are expanded around the normal operation points through the prediction uncertainty analysis of forward and inversion model simultaneously. Finally, the designated operation space from LPPLS inversion is applied in an benchmark glycosylation model.

In this article, we address technical difficulties that arise when applying Markov chain Monte Carlo (MCMC) to hierarchical models designed to perform clustering in the space of latent parameters of subject‐wise generative models. Specifically, we focus on the case where the subject‐wise generative model is a dynamic causal model (DCM) for functional magnetic resonance imaging (fMRI) and clusters are defined in terms of effective brain connectivity. While an attractive approach for detecting mechanistically interpretable subgroups in heterogeneous populations, inverting such a hierarchical model represents a particularly challenging case, since DCM is often characterized by high posterior correlations between its parameters. In this context, standard MCMC schemes exhibit poor performance and extremely slow convergence. In this article, we investigate the properties of hierarchical clustering which lead to the observed failure of standard MCMC schemes and propose a solution designed to improve convergence but preserve computational complexity. Specifically, we introduce a class of proposal distributions which aims to capture the interdependencies between the parameters of the clustering and subject‐wise generative models and helps to reduce random walk behaviour of the MCMC scheme. Critically, these proposal distributions only introduce a single hyperparameter that needs to be tuned to achieve good performance. For validation, we apply our proposed solution to synthetic and real‐world datasets and also compare it, in terms of computational complexity and performance, to Hamiltonian Monte Carlo (HMC), a state‐of‐the‐art Monte Carlo technique. Our results indicate that, for the specific application domain considered here, our proposed solution shows good convergence performance and superior runtime compared to HMC. [ABSTRACT FROM AUTHOR]

Techniques of multivariate pattern analysis (MVPA) can be used to decode the discrete experimental condition or a continuous modulator variable from measured brain activity during a particular trial. In functional magnetic resonance imaging (fMRI), trial-wise response amplitudes are sometimes estimated from the measured signal using a general linear model (GLM) with one onset regressor for each trial. When using rapid event-related designs with trials closely spaced in time, those estimates are highly variable and serially correlated due to the temporally extended shape of the hemodynamic response function (HRF). Here, we describe inverse transformed encoding modelling (ITEM), a principled approach of accounting for those serial correlations and decoding from the resulting estimates, at low computational cost and with no loss in statistical power. We use simulated data to show that ITEM outperforms the current standard approach in terms of decoding accuracy and analyze empirical data to demonstrate that ITEM is capable of visual reconstruction from fMRI signals.

Featured Application: This research can potentially be applied in improving the accuracy of clinical skin cancer diagnostics. In this study, we compare six different machine learning methods in the inversion of a stochastic model for light propagation in layered media, and use the inverse models to estimate four parameters of the skin from the simulated data: melanin concentration, hemoglobin volume fraction, and thicknesses of epidermis and dermis. The aim of this study is to determine the best methods for stochastic model inversion in order to improve current methods in skin related cancer diagnostics and in the future develop a non-invasive way to measure the physical parameters of the skin based partially on the results of the study. Of the compared methods, which are convolutional neural network, multi-layer perceptron, lasso, stochastic gradient descent, and linear support vector machine regressors, we find the convolutional neural network to be the most accurate in the inversion task. [ABSTRACT FROM AUTHOR]

Parameter retrieval and model inversion are key problems in remote sensing and Earth observation. Currently, different approximations exist: a direct, yet costly, inversion of radiative transfer models (RTMs); the statistical inversion with in situ data that often results in problems with extrapolation outside the study area; and the most widely adopted hybrid modeling by which statistical models, mostly nonlinear and non-parametric machine learning algorithms, are applied to invert RTM simulations. We will focus on the latter. Among the different existing algorithms, in the last decade kernel based methods, and Gaussian Processes (GPs) in particular, have provided useful and informative solutions to such RTM inversion problems. This is in large part due to the confidence intervals they provide, and their predictive accuracy. However, RTMs are very complex, highly nonlinear, and typically hierarchical models, so that very often a single (shallow) GP model cannot capture complex feature relations for inversion. This motivates the use of deeper hierarchical architectures, while still preserving the desirable properties of GPs. This paper introduces the use of deep Gaussian Processes (DGPs) for bio-geo-physical model inversion. Unlike shallow GP models, DGPs account for complicated (modular, hierarchical) processes, provide an efficient solution that scales well to big datasets, and improve prediction accuracy over their single layer counterpart. In the experimental section, we provide empirical evidence of performance for the estimation of surface temperature and dew point temperature from infrared sounding data, as well as for the prediction of chlorophyll content, inorganic suspended matter, and coloured dissolved matter from multispectral data acquired by the Sentinel-3 OLCI sensor. The presented methodology allows for more expressive forms of GPs in big remote sensing model inversion problems. [ABSTRACT FROM AUTHOR]

We present a model inversion algorithm, CKLEMAP, for data assimilation and parameter estimation in partial differential equation models of physical systems with spatially heterogeneous parameter fields. These fields are approximated using low-dimensional conditional Karhunen-Loève expansions (CKLEs), which are constructed using Gaussian process regression (GPR) models of these fields trained on the parameters' measurements. We then assimilate measurements of the state of the system and compute the maximum a posteriori (MAP) estimate of the CKLE coefficients by solving a nonlinear least-squares problem. When solving this optimization problem, we efficiently compute the Jacobian of the vector objective by exploiting the sparsity structure of the linear system of equations associated with the forward solution of the physics problem. The CKLEMAP method provides better scalability compared to the standard MAP method. In the MAP method, the number of unknowns to be estimated is equal to the number of elements in the numerical forward model. On the other hand, in CKLEMAP, the number of unknowns (CKLE coefficients) is controlled by the smoothness of the parameter field and the number of measurements, and is generally much smaller than the number of discretization nodes, which leads to a significant reduction of computational cost with respect to the standard MAP method. To show this advantage in scalability, we apply CKLEMAP to estimate the transmissivity field in a two-dimensional steady-state subsurface flow model of the Hanford Site by assimilating synthetic measurements of transmissivity and hydraulic head. We find that the execution time of CKLEMAP scales nearly linearly as N 1.33 , where N is the number of discretization nodes, while the execution time of standard MAP scales as N 2.91. The CKLEMAP method improved execution time without sacrificing accuracy when compared to the standard MAP method. • CKLEMAP is an efficient alternative to the maximum a posteriori (MAP) method for parameter estimation. • Parameter fields are represented using conditional Karhunen-Loève expansions. • Acceleration scheme for Jacobian computations exploits the sparsity of the forward problem. • CKLEMAP and MAP scale as N 1.33 and N 2.91, where N is the number of finite volume cells. • CKLEMAP is as accurate as MAP but significantly faster for large-scale parameter estimation problems. [ABSTRACT FROM AUTHOR]

• Seven families of inversion methods for agricultural applications were identified. • Rationales behind applying inverse modeling in agriculture were summarized. • An operational procedure of inversion method selection was proposed. • The procedure considers practitioners' available resources and goals. Inverse modeling (IM) is a valuable tool in agriculture for estimating model parameters that aid in decision-making. It is particularly useful when parameters cannot be directly measured or easily estimated due to logistical constraints in agricultural settings. Unlike other estimation methods, IM combines a mechanistic model with observations of its outputs to derive the parameters of interest, allowing for the integration of various sources of knowledge. The availability of numerous data sources, such as remote sensing and crowdsourcing, with high spatial and temporal resolution, has expanded the potential of IM in agriculture. Practitioners can now incorporate the spatial and temporal footprint of observational data into parameter estimation. However, common IM techniques currently applied in agriculture often struggle to account for effectively spatial and temporal variability. Relevant IM methods that address these challenges are usually isolated within specific developer and user communities and are not well known within the agricultural community. There is a lack of comprehensive reviews focusing on IM methods suitable for handling spatial and temporal data in agriculture. In parallel, the process of conducting IM in agriculture remains under-formalized. Typically, specific IM methods are chosen for specific combinations of models and types of observational data, but the rationale behind their selection is rarely explained in publications. The relationship between IM methods, models, and observational data is unclear, making it overwhelming for new practitioners to choose an appropriate method. This complex problem, along with the diversity of IM methods, has yet to be adequately addressed while taking into account the specificities of agricultural applications. To address these challenges, this review aims to provide a structured classification of IM methods based on the practical needs of new practitioners in agriculture. It examines a wide range of inversion methods applied in agriculture-related domains and covers four key topics: i) the essential elements and general process of IM, ii) the main families of IM methods in agriculture and their characteristics, iii) the circumstances in which practitioners prefer using IM over other approaches, and their motivations, and iv) practical guidance on choosing a method family based on operational criteria. The review aims to help readers develop a clear understanding of the practice of inverse modeling, gain insights into the diversity of IM methods, and make informed choices when selecting a method family for their agricultural applications. [ABSTRACT FROM AUTHOR]