This work proposes an encrypted controller framework for closed-loop control systems with nonlinear dynamics over fully homomorphic encryption (FHE). Unlike differential privacy and output masking, FHE is a cryptographic primitive that provides assumption-based confidentiality guarantees under standard hardness assumptions. We observe that existing encrypted control frameworks remain largely limited to linear open-loop systems, primarily due to two key challenges: rapid ciphertext noise accumulation in feedback loops and the substantial computational overhead of nonlinear operations. In control systems, feedback is essential for real-time error correction, while nonlinear characteristics are critical for accurately modelling complex system behaviours. To address these challenges, we propose ENClose, a novel encrypted control framework that enables low-latency execution of both feedback control and nonlinear function evaluation. Specifically, ENClose introduces a low-latency homomorphic nonlinear computation framework that accelerates functional bootstrapping (FBS) by combining function segmentation with tree-based encrypted selection. This framework not only mitigates noise accumulation in encrypted feedback loops but also significantly improves the efficiency of FBS under high-precision settings, meeting the computational demands of dynamic control systems. Experimental results show that ENClose achieves a 3× to 20× speedup over state-of-the-art encrypted controllers. We validate ENClose through realworld applications, including multi-vehicle formation, spring–mass–damper control, and anomaly recovery, where the results demonstrate high-precision tracking and successful reconvergence after anomalies.

Electrocardiogram (ECG) signals are crucial indicators of various human physiological states. A thorough analysis of these signals is indispensable for applications such as disease prediction, mental stress assessment, and other medical diagnostics. Despite the rapid progress in large language models (LLM) and their demonstrated prowess in natural language understanding, their application in ECG signal analysis remains underexplored. This paper introduces Tackle Electrocardiogram with Large Language Model Effectively (TELL ME), a novel approach that effectively transfers the robust comprehension capabilities of LLM to ECG signal processing. The method employs a front alignment strategy to align ECG modality with text modality via cross-attention mechanism and incorporates critical manual features into prompts to enhance the performances in specific tasks. The proposed solution has been validated across three downstream tasks, namely, quality assessment, ventricular premature beats detection, and denoising reconstruction, consistently achieving state-of-the-art (SOTA) results.

Fully Homomorphic encryption (FHE) enables high-level security but with a heavy computation workload, necessitating software-hardware co-design for aggressive acceleration. Recent works on specialized accelerators for HE evaluation have made significant progress in supporting lightweight RNS-CKKS applications, especially those with high-density in-memory computing techniques. To fulfill higher computational demands for more general applications, this article proposes multicluster HE accelerating system (MCHEAS), an accelerating system comprising multiple in-situ HE processing accelerators, each functioning as a cluster to perform large-parameter RNS-CKKS evaluation collaboratively. MCHEAS features optimization strategies including the synchronous, preemptive swap, square-diagonal, and odd-even index separation. Using these strategies to compile the computation and transmission of number theoretic transform (NTT) coefficients, the method optimizes the intercluster data swaps, a major bottleneck in NTT computations. Evaluations show that under 1 GHz, with different intercluster data transfer bandwidths, our approach accelerates NTT computations by 26.40% to 51.75%. MCHEAS also improves computing unit utilization by 10.30% to 33.97%, with a maximum peak utilization rate of up to 99.62%. MCHEAS achieves 17.63% to 34.67% speedups for HE operations involving NTT, and 15.12% to 30.62% speedups for demonstrated applications, while enhancing the computing units’ utilization by 5.18% to 21.87% during application execution. Furthermore, we compare MCHEAS with SOTA designs under a specific intercluster data transfer bandwidth, achieving up to$81.45\times $their area efficiencies in applications.

In this paper, we develop a model-free and data efficient batch reinforcement learning algorithm for learning control of continuous state-space and discounted-reward Markov decision processes. This algorithm is an approximate value iteration which uses the manifold regularization method to learn feature representations for Q-value function approximation. The learned features can preserve the intrinsic geometry of the state space by learning on collected samples, and thus can improve the quality of the final value function estimate and the learned policy. The effectiveness and efficiency of the proposed scheme is evaluated on a benchmark control task, i.e., the inverted pendulum balancing problem.

Many areas in China, e.g., Beijing-Tianjin-Hebei area, have been suffering from severe air pollution problem in recent years. The haze has jeopardized people's health and aroused deep concern from the public. To understand residents concerns better and leverage the power of the crowd, the government encourages organizations and individuals to participate in environmental protection activities. In this paper, we describe a decision supporting system with a set of algorithms implemented to extract insights from fused physical data (Air Quality Index, AQI) and social data (social buzz). The system provides insights for the government to closely improve its public services to curb the air pollution, including understanding the correlation between AQI and residents sentiment, spot and remove illegal sources of air pollution.

As the rapid development of Internet of things, a large number of devices, instruments and sensors will connect and collaborate to achieve Smart Planet. One of the key challenges is to integrate devices into business process. SOA is an ideal infrastructure for business process management as applications are invoked using standard interfaces and protocols. It's convenient to use device-oriented web services (doWS) to encapsulate devices functions. A doWS may conflict with other doWS because devices can't be controlled by more than one client at the same time. This brings additional complexity to web service composition. In this paper, we propose a comprehensive device collaboration model to doWS choreography. This model includes device layer, doWS layer, compatible resource layer and process layer, where device layer represent devices, doWS layer encapsulates devices functions, compatible resource layer abstracts compatible device operations as resource and avoid of conflict, process layer describes the collaboration process. Analysis shows that, this model can integrate devices with Web Services, and can achieve the conflict detection of doWS. This model is effective in the Internet of things.