Computing Power Networks (CPNs) present unique scheduling challenges due to the tight coupling of compute and network resources, exacerbated by highly heterogeneous task requirements and significant intra-cycle resource state fluctuations. Addressing these core CPN-specific issues, this paper proposes JARS (Joint Adaptive compute Routing and Scheduling), an intelligent scheduling method based on resource state evolution and heterogeneous task awareness. JARS first identifies the distinct characteristics of heterogeneous tasks—such as compute-intensive, data-intensive, and latency-sensitive types—through dynamic clustering. Crucially, it introduces a real-time resource state evolution mechanism that captures the immediate impact of each scheduling decision within a single cycle, thus overcoming the state perception lag inherent in traditional batch schedulers. We conducted extensive experimental evaluations on six real-world network topologies using a well-defined synthetic workload model that emulates diverse CPN task mixes. Results show that JARS significantly outperforms existing methods, achieving a$\text{1. 8 \%} - \text{2 4 9. 9 \%}$improvement in task success rate and a 9.0%$-84.7 \%$reduction in average completion time, demonstrating its effectiveness and robustness for CPN environments.

The extensive deployment of drone swarms in emergency response and logistics applications presents stringent requirements for mobile ad-hoc network protocols, particularly due to their large-scale and highly dynamic characteristics. As a widely adopted protocol in industrial applications, BATMAN-ADV (Better Approach To Mobile Ad-Hoc Networking Advanced) employs a fixed-interval transmission mechanism for control packets, which leads to inefficient utilization of bandwidth resources. Therefore, we propose ROBIN, a distributed Multi-Agent Reinforcement Learning(MARL) optimization method. Our approach enables each drone to act as an independent agent, autonomously adjusting the control packet transmission interval and optimizing transmission strategies based on localized network states. The solution maintains the protocol’s inherent millisecond end-to-end latency while establishing a dynamic balance between bandwidth consumption and latency performance. Moreover, our method features extremely low resource consumption, making it suitable for drone with limited resources. Experimental results show that ROBIN reduces control packet overhead by 75.9% across diverse network scales and mobility scenarios, while maintaining only 2.29% CPU utilization and 0.25% memory consumption per node on commercial drones.

Strategy card game is a well-known genre that is demanding on the intelligent game-play and can be an ideal test-bench for AI. Previous work combines an end-to-end policy function and an optimistic smooth fictitious play, which shows promising performances on the strategy card game Legend of Code and Magic. In this work, we apply such algorithms to Hearthstone, a famous commercial game that is more complicated in game rules and mechanisms. We further propose several improved techniques and consequently achieve significant progress. For a machine-vs-human test we invite a Hearthstone streamer whose best rank was top 10 of the official league in China region that is estimated to be of millions of players. Our models defeat the human player in all Best-of-5 tournaments of full games (including both deck building and battle), showing a strong capability of decision making.

With the rapid development of artificial intelligence and industrial Internet of Things (IIoT) technologies, intelligent predictive maintenance (IPdM) has received considerable attention from researchers and practitioners. To efficiently predict impending failures and mitigate unexpected downtime, while satisfying the instant maintenance demands of industrial facilities is very important for improving the production efficiency. In this article, a self-attention based “Perception and Prediction” framework, called DeepHealth, is proposed for the instant IPdM. Specifically, the framework is composed of two submodels (i.e., DH-1 and DH-2), which are respectively utilized to perform the health perception and sequence prediction. By operating the framework, the proposed models can predict the health conditions via predicting the future signal samples, thereby completing the instant IPdM. Considering the potential temporal correlation in time series, we deploy an enhanced attention mechanism to capture global dependencies from the vibration signals, and leverage the long- and short-term sequence prediction of sensor signals to support instant maintenance decision-making. On this basis, we conduct a destructive experiment based on the IIoT-enabled rotating machinery and construct a balanced industrial dataset for model evaluations. Extensive experiment results show that the proposed solution achieves good prediction accuracy for instant IPdM on the automatic washing equipment and Case Western Reserve University datasets.

We propose an efficient enhanced K-means clustering (E-KMC) algorithm for semi-blind channel estimation of uplink cell-free massive multiple-input multiple-output (MIMO) systems in factory automation, an important application of the internet of things (IoT). The proposed E-KMC algorithm operates with significantly less clusters and complexity than the KMC algorithm while achieving enhanced bit error rate (BER) performance, as the latter converges extremely slowly even with just medium modulation order and a medium number of transmit antennas. A near-optimal short pilot is designed to assist clustering of the E-KMC based channel estimation scheme. The semi-blind receiver structure achieves a BER performance that is very close to the case with perfect channel state information (CSI), as well as a mean square error (MSE) of channel estimation that is very close to the theoretical lower bound derived in the paper. The proposed E-KMC based channel estimation scheme also significantly outperforms other types of semi-blind channel estimation approaches including second-and higher-order statistics based and machine learning based approaches, while at a much lower complexity. In addition, the E-KMC based channel estimation, is conducted at central processing unit (CPU) and avoids excessive fronthaul overhead due to exchange of the estimated CSI between access points (APs) and CPU.

Speech classification methods mainly focus on the content of the voice segment. To help better underestand the information in a segmented voice, the contents of other segments in the same paragraph should also be paid attention to. In our custom service speech classification problem, we are facing a problem of classification a series of voice segments in a conversation separately into category "custom" or "custom service". Sometimes the voice of both parties in the same conversation can be both sound like a "custom service" or both sound like "custom". In order to make the right prediction, the model needs to know not only the content of the voice segment that it's classifying, but both parties' voice in a conversation, the extra information can help the model to determine who is "more likely" to be a custom service in a conversation. We propose a method called I-CNN, which combines the info-feed layer with CNN. The Info-feed layer allows the CNN to use information from other samples in the same batch, which is helpful in improving the model's performance in our custom service speech classification problem.