Search Results (287)

Whether industrial artificial intelligence (industrial AI) contributes to environmental sustainability remains an open empirical and theoretical question. While digital and intelligent technologies are widely promoted as drivers of green transformation, their net impact on carbon emissions is ambiguous due to potentially offsetting efficiency gains and rebound effects. This study examines how industrial AI influences urban carbon emissions using panel data for 260 Chinese cities from 2005 to 2019. We construct a novel city-level industrial AI development index by integrating information on data infrastructure, AI-related talent supply and intelligent manufacturing services using the entropy weight method. Employing two-way fixed-effects models, instrumental-variable estimations, lag structures, and multiple robustness checks, we identify the causal impact of industrial AI on carbon emissions. The results indicate that industrial AI significantly reduces urban carbon emissions. Mechanism analyses suggest that this effect operates primarily through improvements in energy efficiency and green technological innovation, while being partially offset by scale expansion. Furthermore, a higher share of secondary industry mitigates the emission-reducing effect of industrial AI. Heterogeneity analysis further indicates stronger emission-reduction effects in eastern regions, large cities, and areas with higher human capital and stronger environmental regulation. The findings suggest that intelligent industrial upgrading can simultaneously enhance productivity and support climate mitigation, but this effect is highly context-dependent, offering policy insights for achieving sustainable industrial modernization and carbon neutrality in emerging economies. Full article

3D multi-object tracking (MOT) for autonomous driving remains challenging due to frequent identity switches in crowded scenes, trajectory fragmentation during occlusions, and the difficulty of adapting association strategies to varying scene complexities. While existing methods rely on fixed geometric or appearance-based associations, they struggle to handle ambiguous cases and detection failures. We present an adaptive multi-level 3D MOT framework that achieves robust tracking through three key innovations: (1) multi-granularity temporal modeling that captures both fine-grained short-term motion and coarse long-term trends via dual-scale spatio-temporal attention, enabling accurate motion prediction across different object dynamics; (2) Transformer-based Appearance Association that employs cross-attention to model global inter-object relationships, resolving ambiguous associations in crowded scenarios where geometric cues alone fail; and (3) scene-adaptive learned thresholds that automatically adjust association strictness based on object density, motion complexity, and occlusion levels, avoiding the one-size-fits-all limitations of fixed thresholds. Our hierarchical four-level tracking strategy progressively handles cases from easy geometric matching (Level 1) to complex interval-frame recovery (Level 4), with SOT-based virtual detection generation bridging detector failures. Extensive experiments on the nuScenes benchmark demonstrate state-of-the-art performance. Full article

This paper presents a low-cost retrofitting pipeline for a legacy industrial robot that uses a single RGB webcam and monocular 2D keypoint tracking to estimate human-arm posture angles ${\mathit{\theta }}^{\left(\mathrm{h}\right)}$ and map them to robot-axis joint targets ${\mathbf{q}}_{\mathrm{cmd}}^{\left(\mathrm{r}\right)}$ for A1–A3 on a KUKA KR5-2 ARC HW, while keeping the wrist orientation (A4–A6) fixed. Rather than targeting full six-DoF manipulation, the main contribution is an experimental characterization of how far monocular 2D posture-to-axis mapping can be used reliably for coarse placement and safeguarded low-speed demonstrations on a legacy robot platform. Vision-side accuracy was evaluated per axis against goniometer-based reference angles ${\mathit{\theta }}_{\mathrm{ref}}^{\left(\mathrm{h}\right)}$, showing low errors for A2–A3 within the tested range and larger errors for A1 due to monocular yaw/depth ambiguity and occlusions. The study also analyzes failure modes during simultaneous multi-joint motion, where performance degrades notably, especially for A2 and A3, and reports practical mitigation directions such as improved viewpoints, multi-view/depth sensing, and stricter dropout handling. Runtime behavior is additionally characterized through a loop timing budget, with an end-to-end latency of 185.44 ms and an effective loop frequency of 5.39 Hz, which is consistent with low-speed online operation within the demonstrated scope. The system was implemented in a fenced industrial cell with restricted access and emergency stop; no collaborative operation is claimed. Full article

Wildfires are among the key drivers of transformation in boreal ecosystems; however, the mechanisms of postfire recovery in the arid regions of Eurasia remain insufficiently understood. The aim of this study was to identify the role of burn severity and associated edaphic and hydrological factors in shaping the natural regeneration trajectories of Scots pine forests in the Kostanay region of northern Kazakhstan. This study is based on the integration of field data on seedling regeneration and soil conditions with the analysis of long-term satellite-derived indices (NDVI, NDMI, and NBR). Sample plots were grouped according to fixed burn severity classes, which enabled a consistent statistical comparison and reduced the interpretative ambiguity that has characterized previous studies in the region. The results indicate that pine forest regeneration is most successful under low and moderate burn severity, where seed sources are preserved and favourable moisture conditions are maintained. In contrast, high burn severity leads to a reduction in regenerative potential and a shift in recovery trajectories toward deciduous or sparsely vegetated communities. The spectral indices derived from the remote sensing data strongly agreed with the field-based indicators, confirming their suitability for assessing postfire vegetation dynamics. Soil properties act as important modifying factors of recovery processes, particularly under conditions of limited water availability. These findings enhance the current understanding of postfire recovery mechanisms in the arid part of the boreal zone and highlight the need for differentiated management of postfire landscapes. The integration of field observations with remote sensing data provides a robust framework for monitoring and forecasting recovery processes under an increasingly intensified fire regime. Full article

We develop a categorical framework for modeling recursive uncertainty over preferences in decision theory. Classical models of ambiguity allow for uncertainty over outcomes or beliefs but usually rely on finite or exogenously truncated representations when agents face uncertainty about their own evaluative criteria. Given that such recursive preference formation generates an infinite hierarchy that may not stabilize at any finite level, we introduce a contractive von Neumann–Morgenstern utility functor on a category of compact metric spaces enriched over complete metric spaces, and establish the existence and uniqueness of its canonical fixed point. This fixed point is interpreted as a universal preference space that contains all levels of recursive ambiguity in a consistent and metrically stable form. We further extend the construction to multi-utility representations and discuss its relation to existing models of ambiguity and universal choice spaces. This framework offers a minimal unified representation of recursive preference structures. Full article

Carrier-phase ambiguity validation is essential to ensure the reliability of integer ambiguity resolution in high-precision GNSS positioning. Although integer equivariant (IE) estimators provide optimal integer candidates within their class, noise and model limitations may lead to incorrect fixing. Validation procedures are therefore crucial for safeguarding the transition from float to fixed solutions, particularly in high-precision and safety-critical applications. In this contribution we introduce the concept of Fourier ambiguity validation and show how it is rooted in the principles of integer aperture (IA) estimation and its periodic representation. Unlike classical integer estimators that always return an integer solution, IA estimators introduce adjustable acceptance regions in the float ambiguity domain and fix ambiguities only when sufficient statistical evidence is present. As a result we present a general Fourier representation of IA estimators and provide an analytical description of the probabilistic properties of integer-aperture bootstrapping. We also present a hybrid description and show how the spatial and frequency representations can be mixed so as to do justice to the practical situation when carrier-phase ambiguities have a wide range of varying precision. Full article

In challenging environments, there often exist problems of false alarms and missed detections in real-time kinematic (RTK) ambiguity resolution, which significantly reduce the reliability and availability of position information. To address these issues, a machine-learning method is proposed to conduct a correctness check on RTK ambiguity fixing, aiming to reduce the occurrences of false alarms and missed detections. The inter-system differential RTK model is adopted. Compared with the traditional RTK model, this model can provide an effective feature, namely the differential inter-system biases (DISB), to improve the accuracy of machine-learning classification. This is because when the RTK ambiguity is correctly fixed, the DISB usually appears as a stable constant. In addition to DISB, features that are strongly related to ambiguity fixing, such as the ratio value, DOP value, and residuals, are also comprehensively utilized. This method is verified by an open-source, large-scale, and diverse GNSS/SINS dataset—SmartPNT-POS. The experimental results show that, compared with the traditional method of relying solely on the empirical ratio value for ambiguity fixing verification, the missed detection probability of this method is reduced by 2%, the false-alarm probability is decreased by 29%, and the positioning accuracy is improved by approximately 7%. Moreover, compared with other features, the DISB feature provides the highest contribution rate in the machine-learning classification model. Full article

The monitoring of harmful algal blooms (HABs) constitutes a vital component of marine environmental protection and the sustainable development of the marine economy. However, the highly dynamic nature of these small targets, compounded by the complex water color interference prevalent in the coastal waters where HABs frequently occur, has resulted in traditional remote sensing monitoring methods, particularly those relying on fixed spectral index thresholds and pixel-wise binarization, suffering from imprecise identification in turbid coastal waters where suspended sediments, cloud cover, and sun glint create spectral confusion. These methods also exhibit low automation due to manual threshold adjustment requirements and poor transferability across different spatiotemporal conditions. Consequently, these methods struggle to meet practical application requirements. This study establishes a U-net model-based remote sensing identification framework for red tides using HY-1D CZI imagery (50 m resolution, 1–3 day revisit), targeted negative sample strategies, and event-level accuracy validation methods to achieve efficient marine red tide detection. Targeted negative sample selection involves purposefully selecting spectrally ambiguous regions as negative samples, aiming to enhance recognition accuracy and sample selection efficiency. The combination of targeted sampling with deep learning enables portability to new spatiotemporal contexts by learning invariant spectral–spatial features rather than relying on scene-specific thresholds. Experimental results demonstrate that the targeted negative sample strategy reduces event-level model false negatives by 27%, false positives by 36%, and increases the F1 score by 0.3217. Using an identical sample size, the targeted sample selection strategy yields an F1 score 0.0479 higher than random sampling. To achieve equivalent recognition accuracy, an increased number of random samples would be required. Comparative experiments reveal that the proposed method enhances sample selection efficiency by 87.5%. Transferability is demonstrated through successful identification of red tide patches in Wenzhou waters on 13 April 2022, without model retraining. This demonstrates that red tide remote sensing recognition based on targeted sample selection enables efficient, precise, and automated identification without human intervention, providing a reliable technical solution for operational marine red tide monitoring. Full article

Most existing deep learning models for remote sensing scene classification (RSSC) adopt an offline learning paradigm, where all classes are jointly optimized on fixed-class datasets. In dynamic real-world scenarios with streaming data and emerging classes, such paradigms are inherently prone to catastrophic forgetting when models are incrementally trained on new data. Recently, a growing number of class-incremental learning (CIL) methods have been proposed to tackle these issues, some of which achieve promising performance by rehearsing training data from previous tasks. However, implementing such strategy in real-world scenarios is often challenging, as the requirement to store historical data frequently conflicts with strict memory constraints and data privacy protocols. To address these challenges, we propose a novel memory-efficient feature-replay CIL framework (FR-CIL) for RSSC that retains compact feature embeddings, rather than raw images, as exemplars for previously learned classes. Specifically, a progressive multi-scale feature enhancement (PMFE) module is proposed to alleviate representation ambiguity. It adopts a progressive construction scheme to enable fine-grained and interactive feature enhancement, thereby improving the model’s representation capability for remote sensing scenes. Then, a specialized feature calibration network (FCN) is trained in a transductive learning paradigm with manifold consistency regularization to adapt stored feature descriptors to the updated feature space, thereby effectively compensating for feature space drift and enabling a unified classifier. Following feature calibration, a bias rectification (BR) strategy is employed to mitigate prediction bias by exclusively optimizing the classifier on a balanced exemplar set. As a result, this memory-efficient CIL framework not only addresses data privacy concerns but also mitigates representation drift and classifier bias. Extensive experiments on public datasets demonstrate the effectiveness and robustness of the proposed method. Notably, FR-CIL outperforms the leading state-of-the-art CIL methods in mean accuracy by margins of 3.75%, 3.09%, and 2.82% on the six-task AID, seven-task RSI-CB256, and nine-task NWPU-45 datasets, respectively. At the same time, it reduces memory storage requirements by over 94.7%, highlighting its strong potential for real-world RSSC applications under strict memory constraints. Full article

We propose Adaptive Inference, a portfolio management framework extending Active Inference to non-stationary financial environments. The framework integrates inference, control, and execution under endogenous uncertainty, modeling investment decisions as coupled dynamics of belief updating, preference encoding, and action selection rather than optimization over fixed objectives. In this approach, portfolio behavior is governed by the expected free energy (EFE) minimization, showing that classical valuation models emerge as limiting cases when epistemic components vanish. Using train–test evaluation on the ARKK Innovation ETF (2015–2025), we identify a Passivity Paradox: frozen belief transfer outperforms naive adaptive learning. A Professional Agent achieves a Sharpe ratio of 0.39 while its adaptive counterpart degrades to $-0.28$, reflecting belief contamination when learning from policy-dependent signals. Crucially, the architecture is not designed to generate alpha but to perform endogenous risk management that mitigates overtrading under regime ambiguity and distributional shift. Adaptive Inference Agents maintain long exposure most of the time while tactically reducing positions during high-entropy periods, implementing uncertainty-aware passive investing. All agents reduce realized volatility relative to ARKK Buy-and-Hold (43.0% annualized). Cross-asset validation on the S&P 500 ETF (SPY) shows that inference-guided risk shaping achieves a positive Entropic Sharpe Ratio (ESR), defined as excess return per unit of informational work, thereby quantifying the economic value of information under thermodynamic constraints on inference. Full article

Indoor 3D point cloud semantic segmentation is a fundamental task for fine-grained scene understanding and intelligent perception. Due to the prohibitive cost of dense point-wise annotations, weakly supervised learning has emerged as a promising alternative for indoor point cloud segmentation. However, existing weakly supervised methods commonly rely on fixed confidence thresholds for pseudo-label selection, which exhibit limited generalization caused by threshold sensitivity, underutilization of informative low-confidence regions, and progressive noise accumulation during self-training. To address these issues, we propose TGR-T, a weakly supervised framework for indoor 3D point cloud semantic segmentation that incorporates truncated-Gaussian-weighted reliability with adaptive dynamic thresholding. Specifically, a reliability-adaptive dynamic thresholding strategy is introduced to guide pseudo-label selection based on the evolving confidence statistics of unlabeled mini-batches, with exponential moving average smoothing employed to produce stable global estimates and robust separation of reliable and ambiguous regions. To further exploit uncertain regions, a learnable truncated Gaussian weighting function is designed to explicitly model prediction uncertainty within the ambiguous set, providing soft supervision by assigning adaptive weights to low-confidence predictions during optimization. Extensive experimental results demonstrate that the proposed framework significantly enhances the exploitation of unlabeled data under extremely limited supervision: extensive experiments conducted on standard indoor 3D scene benchmarks demonstrate that TGR-T achieves competitive or superior segmentation performance under extremely sparse supervision and can even outperform several fully supervised baselines trained with dense annotations while using only 1% labeled points, thereby substantially narrowing the performance gap between weakly supervised and fully supervised 3D semantic segmentation methods. Full article

Causal inference in text data has been a demanding objective in the field of natural language processing, mainly due to the intrinsic ambiguity and context sensitivity inherent in data, inducing uncertainty. Diminishing this uncertainty is essential in identifying reliable causal connections and advancing predictive consistency. In this research, we introduce an uncertainty-aware ensemble architecture that combines multiple text embedding schemes with both linear and nonlinear classifiers to boost causal text detection. Both sparse and neural-level embeddings were employed, and then combined it with an ensemble weighting approach based on two uncertainty estimation techniques, namely entropy-based and KL divergence-based. Unlike conventional ensemble methods with uniform or fixed voting strategies, our approach assigns weights inversely proportional to classifier uncertainty, ensuring that confident models exert greater influence on the final decisions. Our results show that TF-IDF, through its effective word frequency weighting scheme, consistently outperforms other embedding techniques, achieving better performance across both linear and nonlinear classifiers on both datasets (News Corpus and CausalLM–Adjective group). The experimental results show that our uncertainty-aware ensemble approach enhances both calibration and confidence predictions. Entropy-based weighting improves confidence in the case of linear classifiers with accuracy, F1-score, entropy and prediction confidence values of 94.3%, 94.0%, 0.382 and 0.774, respectively, while in the case of nonlinear classifiers the KL divergence-based weighting acquires a better performance with an accuracy of 97.6%, F1-score of 97.2%, KL Mean value of around 0.055 and LogLoss of 0.221. Full article

Spacecraft pursuit–evasion in contested environments is complicated by strategic incompleteness: the evader can switch maneuvering modes and deploy multi-domain countermeasures that degrade the pursuer’s perception, leading to non-stationary information and distributionally ambiguous interference statistics. A dynamic time-window Nash equilibrium framework is developed for linearized Local Vertical Local Horizontal (LVLH) relative motion under interference-induced uncertainty. Perceptual degradation is modeled via an evidence–theoretic belief representation, and the Jensen–Shannon (JS) divergence is introduced to quantify discrepancies between nominal and interference-corrupted beliefs. The divergence metric drives an adaptive time-window partitioning policy and an uncertainty-aware running cost that balances nominal performance objectives with robustness regularization during high-degradation intervals. In each time window, sufficient conditions are provided for the existence of a local Nash equilibrium, and equilibrium strategies are characterized by the Hamilton–Jacobi–Bellman–Isaacs (HJBI) equation. A global consistency result is established: assuming state continuity, additive cost decomposition, and dynamic-programming compatibility at window boundaries, concatenating the window-wise equilibria yields a Nash equilibrium over the entire horizon. Unlike conventional receding-horizon differential games with a fixed replanning grid, the proposed policy partitions the horizon online in response to perceptual-degradation events and stitches adjacent windows through a continuation value. This boundary stitching enables the global consistency guarantee under additive costs and state continuity. To hedge against ambiguity in interference intensity, a variational distributionally robust optimization (DRO) problem with moment-constrained ambiguity sets is formulated, and the dual worst-case distribution is derived. The resulting Karush–Kuhn–Tucker (KKT) system is reformulated as a finite-dimensional variational inequality, for which an accelerated Alternating Direction Method of Multipliers (ADMM) operator-splitting solver is proposed for efficient real-time computation. Numerical simulations validate the framework and demonstrate improved robustness and computational scalability under time-varying interference compared with fixed-window baselines. Full article

The dynamic approach integrates Global Positioning System and K-band range-rate (KRR) observations to enable precise orbit determination (POD) and gravity field recovery. However, background model uncertainties and temporal aliasing introduce frequency-dependent noise into the post-fit KRR residuals, thereby degrading overall solution accuracy. To mitigate these effects, empirical signals are typically modeled using either dynamic (DYN) or kinematic (KIN) parameterization strategies. Nevertheless, the combined use of DYN and KIN parameterizations remains largely unassessed, and their potential synergistic impact on POD and gravity field recovery merits systematic evaluation. This study evaluates the individual and joint impacts of DYN and KIN (DYN+KIN) on The Gravity Recovery and Climate Experiment (GRACE) Follow-On orbit accuracy and monthly gravity field recovery using nearly one year of 2019 data (excluding February due to severe data gaps). The refined solutions act as empirical temporal filters, effectively suppressing low-frequency components in KRR residuals, particularly below 1-cycle-per-revolution. Relative to nominal ambiguity-fixed reduced-dynamic orbits, the refined solutions mainly enhance the cross-track component, with DYN+KIN showing the largest improvement, while along-track precision experiences only minor (sub-millimeter) degradation. Overall three-dimensional orbit accuracy improves from 3.8 cm to 3.0 cm (DYN), 2.8 cm (KIN), and 2.8 cm (DYN+KIN). In terms of gravity field recovery, the DYN+KIN solution begins to exhibit more pronounced deviations from the other solutions beyond degree and order 30. Over oceanic regions, residual mass anomaly analysis shows that the DYN+KIN solution is associated with an approximately 16% higher noise level compared to the individual DYN and KIN strategies, which exhibit modest noise reductions relative to the nominal solution. The DYN+KIN also exhibits a dampened ~160-day periodicity in the temporal evolution of low-degree coefficients (e.g., C_2,0), likely due to spectral overlap between empirical parameter frequencies and low-degree gravity signal components. These results indicate that over-parameterization introduces spectral redundancy and absorbs geophysical signals, underscoring the need to balance parameter flexibility and signal fidelity in gravity recovery strategies. Full article

In the era of VUCA (Volatility, Uncertainty, Complexity, Ambiguity), cultivating and enhancing the resilience of SRDI (specialized, refined, distinctive, and innovative) enterprises is critical. Based on existing research, this paper defines enterprise resilience at the beginning and constructs an enterprise resilience evaluation index system that includes three segmented capabilities: recognition and resistance, adaptation and adjustment, and recovery and rebound. Taking 1422 SRDI enterprises in China between 2016 and 2023 as samples, this study conducts an empirical study on the relationship between human capital investment, ambidextrous innovation and SRDI enterprises’ resilience by comprehensively employing various econometric methods such as fixed-effects models, mediating effect tests, and threshold regression. The empirical findings demonstrate that human capital investment positively affects the resilience of SRDI enterprises, with breakthrough and progressive innovation serving as mediating factors. Further research reveals that as scale expands, human capital investment exerts an increasingly strong positive influence on enterprise resilience; concurrently, as resilience improves, the impact of human capital investment shifts from negative to positive, with its positive effect growing progressively stronger. Moreover, increased investment in human capital has a significant positive impact on the recognition and resistance capability, as well as the adaptation and adjustment capability, of SRDI enterprises, but has no significant effect on their recovery and rebound capability. Meanwhile, a heterogeneity analysis by certification type of SRDI enterprises reveals that human capital investment has no significant impact on overall enterprise resilience or on the segmented capabilities of the SRDI “Little Giants” Enterprises. However, for SRDI SMEs, it positively influences their overall enterprise resilience, recognition and resistance capability, and adaptation and adjustment capability. Additionally, for innovative SMEs, it positively impacts their overall enterprise resilience and their recognition and resistance capability. Based on the above, this paper recommends that SRDI enterprises should strengthen their strategic focus, continuously enhance investments in human capital and ambidextrous innovation, implement differentiated human capital investment strategies, and prioritize the recruitment and development of cutting-edge talents. Full article