Search Results (76)

The opto-mechanical automated manufacturing process, characterized by stringent process constraints, dynamic disturbances, and conflicting optimization objectives, presents significant control challenges for traditional scheduling and control approaches. We formulate the scheduling problem within a closed-loop control paradigm and propose a novel bi-level intelligent control framework integrating Deep Reinforcement Learning (DRL) and Bayesian Optimization (BO). The core of our approach is a bi-level intelligent control framework. An inner DRL agent acts as an adaptive controller, generating control actions (scheduling decisions) by perceiving the system state and learning a near-optimal policy through a carefully designed reward function, while an outer BO loop automatically tunes the DRL’s hyperparameters and reward weights for superior performance. This synergistic BO-DRL mechanism facilitates intelligent and adaptive decision-making. The proposed method is extensively evaluated against standard meta-heuristics, including Genetic Algorithm (GA) and Particle Swarm Optimization (PSO), on a complex 20-jobs × 20-machines flexible job shop scheduling benchmark specific to opto-mechanical automated manufacturing. The experimental results demonstrate that our BO-DRL algorithm significantly outperforms these benchmarks, achieving reductions in makespan of 13.37% and 25.51% compared to GA and PSO, respectively, alongside higher machine utilization and better on-time delivery. Furthermore, the algorithm exhibits enhanced convergence speed, superior robustness under dynamic disruptions (e.g., machine failures, urgent orders), and excellent scalability to larger problem instances. This study confirms that integrating DRL’s perceptual decision-making capability with BO’s efficient parameter optimization yields a powerful and effective solution for intelligent scheduling in high-precision manufacturing environments. Full article

In customized production environments featuring multi-task parallelism, the efficient adaptability of edge intelligent models is essential for ensuring the stable operation of production lines. However, rapidly generating deployable lightweight models under conditions of frequent task changes and constrained hardware resources remains a major challenge for current edge intelligence applications. This paper proposes an adaptive lightweight artificial intelligence (AI) model compression method for edge nodes in customized production lines to overcome the limited transferability and insufficient flexibility of traditional static compression approaches. First, a task requirement analysis model is constructed based on accuracy, latency, and power-consumption demands associated with different production tasks. Then, the hardware information of edge nodes is structurally characterized. Subsequently, a compression-strategy candidate pool is established, and an adaptive decision engine integrating ensemble reinforcement learning (RL) and Bayesian optimization (BO) is introduced. Finally, through an iterative optimization mechanism, compression ratios are dynamically adjusted using real-time feedback of inference latency, memory usage, and recognition accuracy, thereby continuously enhancing model performance in edge environments. Experimental results demonstrate that, in typical object-recognition tasks, the lightweight models generated by the proposed method significantly improve inference efficiency while maintaining high accuracy, outperforming conventional fixed compression strategies and validating the effectiveness of the proposed approach in adaptive capability and edge-deployment performance. Full article

A central challenge in artificial intelligence and cognitive science is identifying a unifying principle that governs inference, learning, and action. Active inference proposes such a principle: the minimization of variational free energy. Advocates of active inference argue that the framework subsumes classical models of optimal behavior—including Bayesian decision theory, resource rationality, optimal control, and reinforcement learning—while also instantiating information-theoretic principles such as rate-distortion theory and maximum entropy. However, the literature outlining these conceptual links remains fragmented, limiting integration across fields. This review develops these connections systematically. We show how these major frameworks admit formal correspondences with expected free energy minimization when expressed in variational form, exposing a shared optimization principle that underlies theories of optimal decision-making and information processing. This synthesis is intended both to orient researchers from other fields who are new to active inference and to clarify foundational assumptions for those already working within the framework. Full article

Background/Objectives: Computational studies using drift diffusion models on go/no-go escape tasks consistently show that individuals with suicidal ideation (SI) preferentially engage in active escape from negative emotional states. This study extends these findings by examining how individuals with SI update beliefs about action–outcome contingencies and uncertainty when trying to escape an aversive state. Methods: Undergraduate students with (n = 58) and without (n = 62) a lifetime history of SI made active (go) or passive (no-go) choices in response to stimuli to escape or avoid an unpleasant state in a laboratory-based negative reinforcement task. A Hierarchical Gaussian Filter (HGF) was used to estimate trial-by-trial trajectories of contingency and volatility beliefs, along with their uncertainties, prediction errors (precision-weighted), and dynamic learning rates, as well as fixed parameters at the person level. Bayesian mixed-effects models were used to examine the relationship between trial number, SI history, trial type, and all two-way interactions on HGF parameters. Results: We did not find an effect of SI history, trial type, or their interactions on perceived volatility of reward contingencies. At the trial level, however, participants with a history of SI developed progressively stronger contingency beliefs while simultaneously perceiving the environment as increasingly stable compared to those without SI experiences. Despite this rigidity, they maintained higher uncertainty during escape trials. Participants with an SI history had higher dynamic learning rates during escape trials compared to those without SI experiences. Conclusions: Individuals with an SI history showed a combination of cognitive inflexibility and hyper-reactivity to prediction errors in escape-related contexts. This combination may help explain difficulties in adapting to changing environments and in regulating responses to stress, both of which are relevant for suicide risk. Full article

The heterogeneity of human driving styles in mixed-traffic environments manifests as divergent decision-making behaviors in complex scenarios like highway merging. By accurately recognizing these driving styles and predicting corresponding trajectories, autonomous vehicles can enhance safety, improve traffic efficiency, and concurrently achieve fuel savings in highway merging scenarios. This paper proposes a novel framework wherein a clustering algorithm first establishes statistical priors of driving styles. These priors are then integrated into a Model Predictive Control (MPC) model that leverages Bayesian inference to generate a probability-aware trajectory prediction. Finally, this predicted trajectory is embedded as a component of the state input to a reinforcement learning agent, which is trained using an Actor–Critic architecture to learn the optimal control policy. Experimental results validate the significant superiority of the proposed framework. Under the most challenging high-density traffic scenarios, our method boosts the evaluation reward by 11.26% and the average speed by 10.08% compared to the baseline Multi-Agent Proximal Policy Optimization (MAPPO) algorithm. This advantage also persists in low-density scenarios, where a steady 10.60% improvement in evaluation reward is achieved. These findings confirm that the proposed integrated approach provides an effective decision-making solution for autonomous vehicles, capable of substantially enhancing interaction safety and traffic efficiency in emerging mixed-traffic environments. Full article

Artificial intelligence (AI) is transforming pharmaceutical science by shifting drug delivery research from empirical experimentation toward predictive, data-driven innovation. This review critically examines the integration of AI across formulation design, smart drug delivery systems (DDSs), and sustainable pharmaceutics, emphasizing its role in accelerating development, enhancing personalization, and promoting environmental responsibility. AI techniques—including machine learning, deep learning, Bayesian optimization, reinforcement learning, and digital twins—enable precise prediction of critical quality attributes, generative discovery of excipients, and closed-loop optimization with minimal experimental input. These tools have demonstrated particular value in polymeric and nano-based systems through their ability to model complex behaviors and to design stimuli-responsive DDS capable of real-time therapeutic adaptation. Furthermore, AI facilitates the transition toward green pharmaceutics by supporting biodegradable material selection, energy-efficient process design, and life-cycle optimization, thereby aligning drug delivery strategies with global sustainability goals. However, challenges persist, including limited data availability, lack of model interpretability, regulatory uncertainty, and the high computational cost of AI systems. Addressing these limitations requires the implementation of FAIR data principles, physics-informed modeling, and ethically grounded regulatory frameworks. Overall, AI serves not as a replacement for human expertise but as a transformative enabler, redefining DDS as intelligent, adaptive, and sustainable platforms for future pharmaceutical development. Compared with previous reviews that have considered AI-based formulation design, smart DDS, and green pharmaceutics separately, this article integrates these strands and proposes a dual-framework roadmap that situates current AI-enabled DDS within a structured life-cycle perspective and highlights key translational gaps. Full article

The incorporation of carbon nanotubes (CNTs) enhances the mechanical properties of cement-based materials by inhibiting micro-crack propagation. Machine learning provides an efficient approach for predicting the compressive strength of CNT-reinforced concrete, yet existing studies often lack important features and rely on less adaptive models. To address these issues, a multi-dimensional database (429 experimental data points) covering 11 factors (including cement mix ratio, CNT morphology, and dispersion process) was constructed. A hierarchical model verification and optimization was conducted: traditional regression models (Multiple Linear Regression, Multiple Polynomial Regression (MPR), Multivariate Adaptive Regression Splines), mainstream model (Support Vector Regression (SVR)), and ensemble learning models (Random Forest, eXtreme Gradient Boosting (XGB), Light Gradient Boosting Machine optimized by Particle Swarm Optimization (PSO)/Bayesian Optimization (BO)) are trained, compared, and evaluated. MPR performs best (test set R² = 0.856) among traditional regression models, while SVR (test set R² = 0.824) is less accurate. The highest accuracy in ensemble models is achieved by the PSO-optimized XGB model, with R² = 0.910 (test set). PSO outperforms BO in optimization precision, while BO is much more efficient. Water–cement ratio, age, and sand–cement ratio are the primary influencing factors for strength. Among CNT parameters, the inner diameter has greater impact than the length and outer diameter. Optimal CNT parameters are CNT–cement mass ratio 0.1–0.3%, inner diameter ≥ 7.132 nm, and length 1–15 μm. Surfactant polycarboxylate can increase strength, while OH⁻ functional groups can decrease it. These findings, integrated into the high-precision PSO-XGB model, provide a powerful tool for optimizing the mix design of CNT-reinforced concrete, accelerating its development and application in the industry. Full article

Zinc oxide (ZnO) nanostructures have been intensively investigated for applications in sensing, photocatalysis, and optoelectronic devices, where functional performance is strongly governed by morphology, crystallinity, and defect structure. Conventional wet-chemical and vapor-phase growth methods often require long processing times or complex chemistries and face reproducibility and compatibility challenges when applied to thin, flexible, or curved metallic substrates. Pulsed high-energy techniques—such as pulsed laser deposition (PLD), high-power impulse magnetron sputtering (HiPIMS), and pulsed laser or plasma processing—offer a versatile alternative, enabling rapid and localized synthesis both from and on Zn-bearing thin shells. These methods create transient nonequilibrium conditions that accelerate oxidation and promote spatially controlled nanostructure formation. This review highlights the emerging integration of artificial intelligence (AI) with pulsed ZnO synthesis on thin metallic substrates, emphasizing standardized data reporting, Bayesian optimization and active learning for efficient parameter exploration, physics-informed and graph-based neural networks for predictive modeling, and reinforcement learning for adaptive process control. By connecting synthesis dynamics with data-driven modeling, the review outlines a path toward predictive and autonomous control of ZnO nanostructure formation. Future perspectives include autonomous experimental workflows, machine-vision-assisted diagnostics, and the extension of AI-guided pulsed synthesis strategies to other functional metal oxide systems. Full article

Timing and time synchronization are critical capabilities of Global Navigation Satellite Systems (GNSSs), but their performance deteriorates significantly in challenging environments like urban canyons and tunnels. To address this issue, this paper proposes the Distributed Sensor Time Synchronization architecture (DSTS), a novel architecture integrating Bayesian filtering with deep reinforcement learning. DSTS utilizes Bayesian filtering to fuse Time-of-Flight (ToF) measurements with Channel Impulse Response features for real-time compensation of non-linear errors and accurate path state prediction. Concurrently, the Deep Deterministic Policy Gradient (DDPG) algorithm trains each node into an intelligent agent that dynamically learns optimal synchronization weights based on local information like neighbor clock stability and link quality. This allows the architecture to adaptively amplify reliable nodes while mitigating the negative effects of unstable peers and adverse channels, ensuring high accuracy and availability. Simulation experiments based on a real-world UWB dataset demonstrate the architecture’s exceptional performance. The Bayesian filtering module effectively mitigates non-linear errors, reducing the standard deviation of ToF measurements in NLOS scenarios by up to 51.6% (over 41.2% consistently) while achieving high path state prediction accuracy (>85% static, >95% simulated dynamic). In simulated dynamic and heterogeneous networks, the DDPG algorithm achieves a synchronization accuracy better than traditional average-consensus algorithms, ultimately reaching a frequency and phase precision of $4\times {10}^{-10}$ and $5\times {10}^{-10}$ s, respectively. Full article

This paper introduces the Bateson Game, a signaling game in which ambiguity over the governing rules of interaction (interpretive frames), rather than asymmetry of information about player types, drives strategic outcomes. We formalize the communication paradox of the “double bind” by defining a class of games where a Receiver acts under uncertainty about the operative frame, while the Sender possesses private information about the true frame, benefits from manipulation, and penalizes attempts at meta-communication (clarification). We prove that the game’s core axioms preclude the existence of a separating Perfect Bayesian Equilibrium. More significantly, we show that under boundedly rational learning dynamics, the Receiver’s beliefs can become locked into one of two pathological states, depending on the structure of the Sender’s incentives. If the Sender’s incentives are cyclical, the system enters a persistent oscillatory state (an “ambiguity trap”). If the Sender’s incentives align with reinforcing a specific belief or if the Sender has a dominant strategy, the system settles into a stable equilibrium (a “certainty trap”), characterized by stable beliefs dictated by the Sender. We present a computational analysis contrasting these outcomes, demonstrating empirically how different parametrizations lead to either trap. The Bateson Game provides a novel game-theoretic foundation for analyzing phenomena such as deceptive AI alignment and institutional gaslighting, demonstrating how ambiguity can be weaponized to create durable, exploitative strategic environments. Full article

Voltage stability in DC microgrids (DC MG) is crucial for ensuring reliable operation and component safety. This paper surveys voltage control techniques for DC MG, classifying them into model-based, model-free, and hybrid approaches. It analyzes their fundamental principles and evaluates their strengths and limitations. In addition to the survey, the study investigates the voltage control problem in a critical scenario involving a DC/DC buck converter with an input LC filter. Two model-free deep reinforcement learning (DRL) control strategies are proposed: twin-delayed deep deterministic policy gradient (TD3) and proximal policy optimization (PPO) agents. Bayesian optimization (BO) is employed to enhance the performance of the agents by tuning their critical hyperparameters. Simulation results demonstrate the effectiveness of the DRL-based approaches: compared to benchmark methods, BO-TD3 achieves the lowest error metrics, reducing root mean square error (RMSE) by up to 5.6%, and mean absolute percentage error (MAPE) by 7.8%. Lastly, the study outlines future research directions for DRL-based voltage control aimed at improving voltage stability in DC MG. Full article

Modern AI systems can strategically misreport information when incentives diverge from truthfulness, posing risks for oversight and deployment. Prior studies often examine this behavior within a single paradigm; systematic, cross-architecture evidence under a unified protocol has been limited. We introduce the Strategy Elicitation Battery (SEB), a standardized probe suite for measuring deceptive reporting across large language models (LLMs), reinforcement-learning agents, vision-only classifiers, multimodal encoders, state-space models, and diffusion models. SEB uses Bayesian inference tasks with persona-controlled instructions, schema-constrained outputs, deterministic decoding where supported, and a probe mix (near-threshold, repeats, neutralized, cross-checks). Estimates use clustered bootstrap intervals, and significance is assessed with a logistic regression by architecture; a mixed-effects analysis is planned once the per-round agent/episode traces are exported. On the latest pre-correction runs, SEB shows a consistent cross-architecture pattern in deception rates: ViT $80.0\%$, CLIP $15.0\%$, Mamba $10.0\%$, RL agents $10.0\%$, Stable Diffusion $10.0\%$, and LLMs $5.0\%$ (20 scenarios/architecture). A logistic regression on per-scenario flags finds a significant overall architecture effect (likelihood-ratio test vs. intercept-only: ${\chi }^2\left(5\right)=41.56$, $p=7.22\times {10}^{-8}$). Holm-adjusted contrasts indicate ViT is significantly higher than all other architectures in this snapshot; the remaining pairs are not significant. Post-correction acceptance decisions are evaluated separately using residual deception and override rates under SEB-Correct. Latency varies by architecture (sub-second to minutes), enabling pre-deployment screening broadly and real-time auditing for low-latency classes. Results indicate that SEB-Detect deception flags are not confined to any one paradigm, that distinct architectures can converge to similar levels under a common interface, and that reporting interfaces and incentive framing are central levers for mitigation. We operationalize “deception” as reward-sensitive misreport flags, and we separate detection from intervention via a correction wrapper (SEB-Correct), supporting principled acceptance decisions for deployment. Full article

The prediction of Remaining Useful Life (RUL) constitutes a vital aspect of Prognostics and Health Management (PHM), providing capabilities for the assessment of mechanical component health status and prediction of failure instances. Recent studies on feature extraction, time-series modeling, and multi-task learning have shown remarkable advancements. However, most deep learning (DL) techniques predominantly focus on unimodal data or static feature extraction techniques, resulting in a lack of RUL prediction methods that can effectively capture the individual differences among heterogeneous sensors and failure modes under complex operational conditions. To overcome these limitations, an adaptive RUL prediction framework named ADAPT-RULNet is proposed for mechanical components, integrating the feature extraction capabilities of attention-enhanced deep learning (DL) and the decision-making abilities of deep reinforcement learning (DRL) to achieve end-to-end optimization from raw data to accurate RUL prediction. Initially, Functional Alignment Resampling (FAR) is employed to generate high-quality functional signals; then, attention-enhanced Dynamic Time Warping (DTW) is leveraged to obtain individual degradation stages. Subsequently, an attention-enhanced of hybrid multi-scale RUL prediction network is constructed to extract both local and global features from multi-format data. Furthermore, the network achieves optimal feature representation by adaptively fusing multi-source features through Bayesian methods. Finally, we innovatively introduce a Deep Deterministic Policy Gradient (DDPG) strategy from DRL to adaptively optimize key parameters in the construction of individual degradation stages and achieve a global balance between model complexity and prediction accuracy. The proposed model was evaluated on aircraft engines and railway freight car wheels. The results indicate that it achieves a lower average Root Mean Square Error (RMSE) and higher accuracy in comparison with current approaches. Moreover, the method shows strong potential for improving prediction accuracy and robustness in varied industrial applications. Full article

Given that urban mobility arises from the interaction between social and environmental conditions, this study constructs a Bayesian network to represent these relationships in David, Panama, using 500 georeferenced household surveys that recorded variables related to demographics, travel behavior, infrastructure, mobility patterns and perceptions of risk, safety, and vulnerability. The Bayesian network was built and validated through a consensus-driven hybrid procedure combining structural learning and expert knowledge, resulting in a directed acyclic graph (DAG) with 127 nodes and 189 arcs; and conditional probability tables (CPTs) were learned from data. The topology of the network was analyzed with Louvain community detection, revealing eleven subsystems that group household economy and mode choice, hydrometeorological mobility barriers, congestion, public-transport quality, and safety in school travel. The inferences show gender-based differences in the risk of harassment on public transport, higher perceived vulnerability on longer trips, and elevated stress among middle-aged drivers. The model highlights potential priority interventions such as reinforcing public-transport safety, promoting self-contained trips, and encouraging short-distance active mobility, based on population perceptions. The resulting DAG functions as both an analytical and communication tool for urban management, is visually understandable to all stakeholders, and provides unprecedented evidence for Panama in a little-studied context. Full article

Clean water security demands autonomous systems that sense, reason, and act at scale. We introduce AquaSurveil, a unified multi-agent platform coupling mobile robots, fixed IoT nodes, and privacy-preserving machine learning for continent-scale water health surveillance. The architecture blends Gaussian-process mapping with distributed particle filtering, multi-agent deep-reinforcement Voronoi coverage, GAN/LSTM anomaly detection, and sheaf-theoretic data fusion; components are tuned by Bayesian optimization and governed by Age-of-Information-aware power control. Evaluated on a 2.82-million-record dataset (1940–2023; five countries), AquaSurveil achieves up to 96% spatial-coverage efficiency, an ROC-AUC of 0.96 for anomaly detection, ≈95% state-estimation accuracy, and reduced energy consumption versus randomized patrols. These results demonstrate scalable, robust, and energy-aware water quality surveillance that unifies robotics, the IoT, and modern AI. Full article