Prediction Of Future States Research Articles

Comparison of reduced order models based on dynamic mode decomposition and deep learning for predicting chaotic flow in a random arrangement of cylinders

Three reduced order models are evaluated in their capacity to predict the future state of an unsteady chaotic flow field. A spatially fully developed flow generated in a random packing of cylinders at a solid fraction of 0.1 and a nominal Reynolds number of 50 is investigated. For deep learning (DL), convolutional autoencoders are used to reduce the high-dimensional data to lower dimensional latent space representations of size 16, which were then used for training the temporal architectures. To predict the future states, two DL based methods, long short-term memory and temporal convolutional neural networks, are used and compared to the linear dynamic mode decomposition (DMD). The predictions are tested in their capability to predict the spatiotemporal variations of velocity and pressure, flow statistics such as root mean squared values, and the capability to predict fluid forces on the cylinders. Relative errors between 15% and 20% are evident in predicting instantaneous velocities, chiefly resulting from phase differences between predictions and ground truth. The spatial distribution of statistical second moments is predicted to be within a maximum of 5%–10% of the ground truth with mean error in the range of 1%–2%. Using the predicted fields, instantaneous fluid drag force predictions on individual particles exhibit a mean relative error within 20%, time-averaged drag force predictions to within 5%, and total drag force over all particles to within 1% of the ground truth values. It is found that overall, the non-linear DL models are more accurate than the linear DMD algorithm for the prediction of future states.

Smart Preventive Maintenance of Hybrid Networks and IoT Systems Using Software Sensing and Future State Prediction.

At present, IoT and intelligent applications are developed on a large scale. However, these types of new applications require stable wireless connectivity with sensors, based on several standards of communication, such as ZigBee, LoRA, nRF, Bluetooth, or cellular (LTE, 5G, etc.). The continuous expansion of these networks and services also comes with the requirement of a stable level of service, which makes the task of maintenance operators more difficult. Therefore, in this research, an integrated solution for the management of preventive maintenance is proposed, employing software-defined sensing for hardware components, applications, and client satisfaction. A specific algorithm for monitoring the levels of services was developed, and an integrated instrument to assist the management of preventive maintenance was proposed, which are based on the network of future states prediction. A case study was also investigated for smart city applications to verify the expandability and flexibility of the approach. The purpose of this research is to improve the efficiency and response time of the preventive maintenance, helping to rapidly recover the required levels of service, thus increasing the resilience of complex systems.

Predictor‐based switching control for networked systems

SummaryThis paper concerns a networked predictive control (NPC) for linear systems with data time‐varying delays, packet losses and disordering. By utilizing the received time‐delayed state measurement and control input, a predictive algorithm is firstly proposed for approximating the future state prediction. Afterwards, a probability dependent switching control law is proposed. Although the large data transmission delays emerge in the forward and backward channels, the packet disordering in the backward channel can be naturally excluded thus the actuator always receives valid control command during each sampling period with correct time sequence. The proposed NPC not only guarantees the global uniform exponential stability of the overall networked system but also brings merits in tolerating locally unstable sub‐systems. Numerical examples are provided to demonstrate the effectiveness and improvements of the proposed method.

Online and offline learning of player objectives from partial observations in dynamic games

Robots deployed to the real world must be able to interact with other agents in their environment. Dynamic game theory provides a powerful mathematical framework for modeling scenarios in which agents have individual objectives and interactions evolve over time. However, a key limitation of such techniques is that they require a priori knowledge of all players’ objectives. In this work, we address this issue by proposing a novel method for learning players’ objectives in continuous dynamic games from noise-corrupted, partial state observations. Our approach learns objectives by coupling the estimation of unknown cost parameters of each player with inference of unobserved states and inputs through Nash equilibrium constraints. By coupling past state estimates with future state predictions, our approach is amenable to simultaneous online learning and prediction in receding horizon fashion. We demonstrate our method in several simulated traffic scenarios in which we recover players’ preferences, for, e.g. desired travel speed and collision-avoidance behavior. Results show that our method reliably estimates game-theoretic models from noise-corrupted data that closely matches ground-truth objectives, consistently outperforming state-of-the-art approaches.

Zooplankton grazing is the largest source of uncertainty for marine carbon cycling in CMIP6 models

The current generation of Earth system models used by the United Nations to project future climate scenarios (CMIP6) relies heavily on marine biogeochemical models to track the fate of carbon absorbed into the oceans. Here we compare 11 CMIP6 marine biogeochemical models and find the largest source of inter-model uncertainty in their representation of the marine carbon cycle is phytoplankton-specific loss rates to zooplankton grazing. This uncertainty is over three times larger than that of net primary production and driven by large differences in prescribed zooplankton grazing dynamics. We run a controlled sensitivity experiment in a global marine biogeochemical model and find that small changes in prescribed grazing dynamics (roughly 5% of what is used across CMIP6 models) can increase secondary and export production by 5 and 2 PgC yr−1, respectively, even when tuned to identical net primary production, likely biasing predictions of future climate states and food security.

Catching up with the trees: empirical advancements to improve herbaceous representation in models.

This article is a Commentary on Curasi et al. (2023), 239: 562–575.

A data-driven predictive maintenance model for hospital HVAC system with machine learning

ABSTRACT Corrective and preventive maintenance strategies are typically employed to maintain an efficient functionality of different facility systems. This entails the evaluation of current conditions and the prediction of future conditions. Such prediction is highly needed for critical building systems such as Heating, Ventilation, and Air Conditioning (HVAC) of hospitals to maintain their functionality and extend their lifetime. Current literature highlights the benefits of adopting machine-learning algorithms for predictive modelling. Literature also reveals a gap in predictive modelling based on real-time sensor data and the prediction of both short-term and long-term future conditions. This paper presents a data-driven predictive maintenance model of a hospital’s HVAC system with a focus on the Air Handling Units (AHUs). The developed model adopts machine-learning using the sensor data acquired by the BMS and the database of the hospital’s CMMS. Support Vector Machine (SVM), Decision Trees (DT), and K-Nearest Neighbours (KNN) algorithms are used for the prediction of AHU’s short-term conditions. Prophet Forecasting and Seasonal Auto-Regressive Integrated Moving Average (SARIMA) algorithms are then used to predict the AHU’s long-term future conditions. The study also highlights the benefits of adopting the proposed model in terms of reduced maintenance cost and improved operational effectiveness of hospital AHUs.

A Cybersecurity Knowledge Graph Completion Method for Penetration Testing

Penetration testing is an effective method of making computers secure. When conducting penetration testing, it is necessary to fully understand the various elements in the cyberspace. Prediction of future cyberspace state through perception and understanding of cyberspace can assist defenders in decision-making and action execution. Accurate cyberspace detection information is the key to ensuring successful penetration testing. However, cyberspace situation awareness still faces the following challenges. Due to the limited detection capability, the information obtained from cyberspace detection intelligence is incomplete. There are some errors in the cyberspace detection intelligence, which may mislead the penetration testing workers. The knowledge graph can store and manage the cybersecurity data. In order to ensure the integrity and accuracy of cyberspace information, we design a knowledge graph completion model called CSNT to complete cybersecurity data. CSNT uses the BiLSTM to capture the interaction information between entities and relationships. It models the relationship between entities by combining the neural network and tensor decomposition. The Pearson Mix Net is designed to control the generation of joint vectors. We also design a novel self-distillation strategy to reduce catastrophic forgetting during model training. After learning the relationship pattern between entities in the cyberspace detection intelligence, the model can be used to mine the knowledge not found in the cybersecurity detection intelligence and correct the erroneous records. Experiments show that our method has certain advantages for the knowledge graph completion.

Generalizable turbulent flow forecasting for adaptive optics control.

In this work, we characterize the capability of artificial neural network predictive models for generalizable turbulence forecasting, particularly for use in predictive adaptive optics (AO) applications. Predictive AO control, which utilizes future state predictions of an optical wavefront propagated through a turbulent medium to drive correction, is a promising technology for optical propagation in high-disturbance and low-signal environments. The dynamics describing the evolution of turbulent flow can vary greatly. Accordingly, a generalizable approach to turbulence forecasting has key benefits in allowing for prediction across a range of conditions, thus enabling continuous predictive AO operation in dynamic environments and having reduced sensitivity to changes in conditions. We present a model for generalizable turbulence forecasting, which demonstrated consistent high performance over a range of compressible flow conditions outside those included in the training sample, with only a minimal increase in prediction error compared with a hypothetical baseline model, which assumes perfect a priori characterization. These results demonstrate a clear ability to extract useful dynamics from a limited domain of turbulent conditions and apply these appropriately for forecasting, which could inform future design of predictive AO systems.

Identifying the true drivers of abrupt changes in ecosystem state with a focus on time lags: Extreme precipitation can determine water quality in shallow lakes

A better understanding of abrupt ecosystem changes is needed to improve prediction of future ecosystem states under climate change. Chronological analysis based on long-term monitoring data is an effective way to estimate the frequency and magnitude of abrupt ecosystem changes. In this study, we used abrupt-change detection to differentiate changes of algal community composition in two Japanese lakes and to identify the causes of long-term ecological transitions. Additionally, we focused on finding statistically significant relationships between abrupt changes to aid with factor analysis. To estimate the strengths of driver–response relationships underlying abrupt algal transitions, the timing of the algal transitions was compared to that of abrupt changes in climate and basin characteristics to identify any synchronicities between them. The timing of abrupt algal changes in the two study lakes corresponded most closely to that of heavy runoff events during the past 30–40 years. This strongly suggests that changes in the frequency of extreme events (e.g., heavy rain, prolonged drought) have a greater effect on lake chemistry and community composition than do shifts in the means of climate and basin factors. Our analysis of synchronicity (with a focus on time lags) could provide an easy method to identify better adaptative strategies for future climate change.

Health Indicator Similarity Analysis-Based Adaptive Degradation Trend Detection for Bearing Time-to-Failure Prediction

Time-to-failure (TTF) prediction of bearings is vital to the prognostic and health management of rotating machines. Owing to the shifty degradation trends (DTs) of bearings, it is still difficult to obtain accurate TTF prognostic results. To solve this problem, this paper proposes an online, continuously updated TTF prognostic method based on health indicator (HI) similarity analysis and DT detection. First, multiple degradation features are extracted and fused to construct principal component HI by using dynamic principal component analysis. Next, exponential degradation models are fitted using the HI values for future state prediction. By regarding several HI values as a tested segment, the DT is detected by analyzing the similarity of the tested segment and the fitted curve. Finally, TTF is predicted by extrapolating the DT to hit the estimated failure threshold. Two case studies based on public bearing datasets demonstrate the superiority of the proposed approach over state-of-the-art methods.

Structural inference of networked dynamical systems with universal differential equations.

Networked dynamical systems are common throughout science in engineering; e.g., biological networks, reaction networks, power systems, and the like. For many such systems, nonlinearity drives populations of identical (or near-identical) units to exhibit a wide range of nontrivial behaviors, such as the emergence of coherent structures (e.g., waves and patterns) or otherwise notable dynamics (e.g., synchrony and chaos). In this work, we seek to infer (i) the intrinsic physics of a base unit of a population, (ii) the underlying graphical structure shared between units, and (iii) the coupling physics of a given networked dynamical system given observations of nodal states. These tasks are formulated around the notion of the Universal Differential Equation, whereby unknown dynamical systems can be approximated with neural networks, mathematical terms known a priori (albeit with unknown parameterizations), or combinations of the two. We demonstrate the value of these inference tasks by investigating not only future state predictions but also the inference of system behavior on varied network topologies. The effectiveness and utility of these methods are shown with their application to canonical networked nonlinear coupled oscillators.

New indicators of ecological resilience and invasion resistance to support prioritization and management in the sagebrush biome, United States

Ecosystem transformations to altered or novel ecological states are accelerating across the globe. Indicators of ecological resilience to disturbance and resistance to invasion can aid in assessing risks and prioritizing areas for conservation and restoration. The sagebrush biome encompasses parts of 11 western states and is experiencing rapid transformations due to human population growth, invasive species, altered disturbance regimes, and climate change. We built on prior use of static soil moisture and temperature regimes to develop new, ecologically relevant and climate responsive indicators of both resilience and resistance. Our new indicators were based on climate and soil water availability variables derived from process-based ecohydrological models that allow predictions of future conditions. We asked: (1) Which variables best indicate resilience and resistance? (2) What are the relationships among the indicator variables and resilience and resistance categories? (3) How do patterns of resilience and resistance vary across the area? We assembled a large database (n = 24,045) of vegetation sample plots from regional monitoring programs and derived multiple climate and soil water availability variables for each plot from ecohydrological simulations. We used USDA Natural Resources Conservation Service National Soils Survey Information, Ecological Site Descriptions, and expert knowledge to develop and assign ecological types and resilience and resistance categories to each plot. We used random forest models to derive a set of 19 climate and water availability variables that best predicted resilience and resistance categories. Our models had relatively high multiclass accuracy (80% for resilience; 75% for resistance). Top indicator variables for both resilience and resistance included mean temperature, coldest month temperature, climatic water deficit, and summer and driest month precipitation. Variable relationships and patterns differed among ecoregions but reflected environmental gradients; low resilience and resistance were indicated by warm and dry conditions with high climatic water deficits, and moderately high to high resilience and resistance were characterized by cooler and moister conditions with low climatic water deficits. The new, ecologically-relevant indicators provide information on the vulnerability of resources and likely success of management actions, and can be used to develop new approaches and tools for prioritizing areas for conservation and restoration actions.

Learning globally linear predictors using deep Koopman embeddings with application to marine vehicles

Linearity of the model for controlled dynamical systems is a very desirable property because of its simplicity in the state prediction and control. Koopman operator theory provides a framework for global mapping of a nonlinear system into an equivalent linear system. The goal of this work is to exploit Koopman theory and modern machine learning techniques to find the linear system representation of the underlying nonlinear system for future state predictions. The model generated in this way is completely data driven and requires no a priori knowledge of the underlying dynamics system. The model is applied to two marine vehicles whose trajectories are generated using simulation and evaluated against common model identification techniques. The results show that proposed method is comparable to conventional identification methods and even outperforms them in cases when complex nonlinear dynamics, which is often neglected, becomes relevant.

Gaussian Radar Transformer for Semantic Segmentation in Noisy Radar Data

Scene understanding is crucial for autonomous robots in dynamic environments for making future state predictions, avoiding collisions, and path planning. Camera and LiDAR perception made tremendous progress in recent years, but face limitations under adverse weather conditions. To leverage the full potential of multi-modal sensor suites, radar sensors are essential for safety critical tasks and are already installed in most new vehicles today. In this paper, we address the problem of semantic segmentation of moving objects in radar point clouds to enhance the perception of the environment with another sensor modality. Instead of aggregating multiple scans to densify the point clouds, we propose a novel approach based on the self-attention mechanism to accurately perform sparse, single-scan segmentation. Our approach, called Gaussian Radar Transformer, includes the newly introduced Gaussian transformer layer, which replaces the softmax normalization by a Gaussian function to decouple the contribution of individual points. To tackle the challenge of the transformer to capture long-range dependencies, we propose our attentive up- and downsampling modules to enlarge the receptive field and capture strong spatial relations. We compare our approach to other state-of-the-art methods on the RadarScenes data set and show superior segmentation quality in diverse environments, even without exploiting temporal information.

The relationship between psychiatric symptoms and affective forecasting bias

Background and objectivesAffective forecasting (AF) is the prediction of future emotional states. Negatively biased affective forecasts (i.e., overestimating negative affect) have been associated with trait anxiety, social anxiety, and depression symptoms, but few studies have tested these relationships while covarying commonly co-occurring symptoms. MethodsIn this study, participants (N = 114) completed a computer game in dyads. Participants were randomized into one of two conditions: a condition in which they were led to believe they were at-fault for losing their dyad money (n = 24 dyads) or a condition in which they were told no one was at fault (n = 34 dyads). Prior to the computer game, participants forecasted their affect for each potential game outcome. ResultsMore severe social anxiety, trait-level anxiety, and depressive symptoms were all associated with more negative AF bias in the at-fault relative to the no-fault condition, and this effect persisted when controlling for other symptoms. Cognitive and social anxiety sensitivity was also associated with more negative AF bias. LimitationsThe generalizability of our findings is innately limited by our non-clinical, undergraduate sample. Future work should replicate and extend our research in more diverse populations and clinical samples. ConclusionsOverall, our results support that AF biases are observed across a range of psychopathology symptoms and associated with transdiagnostic cognitive risk factors. Future work should continue investigating the etiological role of AF bias in psychopathology.

Heads-Up Green in Connected Traffic Signals

Information-sharing between traffic signals and connected vehicles can enhance the traffic conditions in signalized intersections. We propose a straightforward application, heads-up green, with the aim to inform queuing connected vehicles about a switch from red to green. In this way, the drivers’ reaction time is expected to be reduced and the number of vehicles that can pass the intersection during a cycle is thereby increased. Unlike the applications Green Light Optimal Speed Advisory (GLOSA), time-to-green and time-to-red, there is no need for predictions of future traffic signal states. Therefore, heads-up green can be used in traffic actuated and adaptive control strategies in signalized intersections without loss in precision. A simulation-based evaluation of a typical Swedish intersection with a traffic actuated controller is performed to evaluate the traffic performance. The main goal as been to show on the potential to use heads-up green to increase traffic efficiency in signalized intersections. Additionally, different reaction times to the information from heads-up green has been investigated. The results show that with heads-up green it is possible to improve the travel time by up to 15% at high demand levels when queues are frequently observed. The improvements are largest with 100% connected vehicles, but significant improvements of 5-15% are observed also for lower shares of connected vehicles. At high demand levels, travel times are decreased already at small reductions in the reaction time. Hence, there is a potential to improve traffic efficiency already when small changes in reaction time can be achieved with heads-up green.

Subpredictor-Based Fuzzy Control Design for a Reaction–Diffusion Equation

This article addresses the stabilization of the nonlinear reaction–diffusion equation with an unknown but bounded delay. Based on Takagi–Sugeno (T–S) fuzzy rules, a T–S fuzzy partial differential equation (PDE) model is constructed to describe the nonlinear distributed parameter system. More importantly, we contribute to the large delay compensation and the future state prediction by introducing a chain of subpredictors for the T–S fuzzy PDE model, and the convergence of observation errors is guaranteed as well. Different from the conventional approaches, the proposed observer can compensate a larger delay. To stabilize the system, for the first time, a new subpredictor-based T–S fuzzy control law is suggested by using the proposed observer, which is composed of elementary observers connected in series. The implementation of the subpredictor feedback is simpler than the classical methods that only one of the elementary observers is employed. Via Lyapunov–Krasovskii approach, sufficient exponential stability conditions are established in the framework of linear matrix inequalities. A numerical example is provided to illustrate the effectiveness of theoretical results.

Energy management for hybrid electric vehicles based on imitation reinforcement learning

An effective energy management strategy (EMS) in hybrid electric vehicles (HEVs) is indispensable to promote consumption efficiency due to time-varying load conditions. Currently, learning based algorithms have been widely applied in energy controlling performance of HEVs. However, the enormous computation intensity, massive data training and rigid requirement of prediction of future operation state hinder their substantial exploitation. To mitigate these concerns, an imitation reinforcement learning-based algorithm with optimal guidance is proposed in this paper for energy control of hybrid vehicles to accelerate the solving process and meanwhile achieve preferable control performance. Firstly, offline global optimization is firstly conducted considering various driving conditions to search power allocation trajectories. Then, the battery depletion boundaries with respect to driving distance are introduced to generate a narrowed state space, in which the optimal trajectory is fused into the training process of reinforcement learning to guide the high-efficiency strategy production. The simulation validations reveal that the proposed method provides preferable energy reduction for HEVs in arbitrary driving scenarios, and suggests an efficient solution instruction for similar problems in mechanical and electrical systems with constraints and optimal information.

An Interrelated Imitation Learning Method for Heterogeneous Drone Swarm Coordination

The proliferation of small drones has boosted diverse intelligent services, in which the effective swarm coordination plays a vital role in enhancing execution efficiency. However, owing to unreliable air communication and heterogeneous computation capabilities, it is difficult to achieve coordinated actions particularly in distributed scenarios with incomplete observations. In this article, we utilize the generative adversarial imitation learning (GAIL) model to coordinate the drones’ maneuvers by imitating the peer's demonstrations. However, incomplete observations will lead to inaccurate imitation policies. In order to recover true environment states, we encode historical observation-action trajectories into latent belief representations, which are trained in correlation to imitation policies. Moreover, by merging the trace of historical contexts, the prediction of future states and the action-assisted guidance information, we gain robust belief representations, which lead to more accurate imitation policies. We evaluate the algorithm performance via the drones’ formation control task. Experiment results display the superiorities on imitation accuracy, execution time and energy cost.