The computational models for geophysical fluid dynamics are computationally enormously expensive to employ in tasks such as data assimilation and uncertainty quantification. Naturally, surrogate models seeking to alleviate the computational burden has been proposed. Researchers have started applying artificial intelligence and machine learning algorithms, particularly artificial neural networks, to build data-driven surrogate models for geophysical flows. The performance of the neural network highly relies upon their architecture design and selection of hyper-parameters. These neural networks are usually manually crafted through trial and error to maximize their performance. This often demands specialized knowledge of the underlying neural network as well as the domain problems of interest. This limitation can be addressed by using an evolutionary algorithm to automatically design and select optimal hyper-parameters of the neural network. In this study, the genetic algorithm is applied to effectively design the long short-term memory (LSTM) neural network to build the forecasting model of the temperature in the NOAA sea-surface temperature data set.

How does intelligence emerge? What kinds of dynamical behaviors are intertwined with intelligence and how do we control them? Concerning with these two issues, relevant studies on intelligent control are briefly surveyed from an integrated viewpoint of complex network and dynamic systems. Presented first are fundamental concepts and challenging questions that arise from the interdisciplinary research areas of complex networks, dynamical systems, neuroscience, and intelligent control. An overview of research progress on intelligent control is further presented, emphasizing pinning control, hybrid control, adaptive control, and the controllability of complex networks. Moreover, potential applications of complex network dynamics and intelligent control in the fields of brain science and machine behavior are briefly discussed, with an outlook at possible research directions.

Optical coherence tomography (OCT) is an emerging imaging technique that measures internal structures of transparent and semitransparent objects, e.g. polymers, ceramics, and composites, with micrometer resolution and millimeter range. After combining with the method of phase-contrast, the technique is also available for measuring nanometer level displacement field and micro-strain level deformation field inside objects due to the high sensitivity of interference phase. Since it can be employed for characterizing and testing full-field mechanical behaviors of inside materials, the technique has gained rapid development and more attention in the last decade, which has already become a hot topic in non-destructive testing. In this review, the basic principle of phase-contrast OCT was firstly described, some typical applications were then introduced, e.g. depth-resolved thermal deformation measurement of multilayer structure, curing processing visualization inside polymer, and micro defect identification inside materials, and some future developments were finally discussed.

A robust tracking control problem for linear discrete-time systems with a prescribed convergence rate is considered. The robust tracking problem is formulated by utilizing robust output regulation and is subsequently solved by reinforcement learning with integration of the prescribed convergence rate. The learned controller ensures that the tracking error asymptotically converges to zero, meanwhile it is robust to uncertain system dynamics. The proposed convergence rate design is data-driven in the sense that it does not depend on the time for the system evolution or the accurate system model.

Compute first networking (CFN) is a latest distributed framework that intelligently allocates computing resources for edge computing according to computing load and network status. It requires real-time visibility of available statuses of local or remote computing resources. To the best of our knowledge, thisis the first endeavor to propose a centralized fault-detection protocol called CFN-Watchdog to well meet this CFN requirement and timely recycle resources occupied by faults. The impact of various parameters (e.g., detection thresholds, task processing time, and network delay) on the Watchdog performance is then theoretically analyzed. Extensive simulations verify the effectiveness of our proposed protocol and the accuracy of our theoretical model. This study is very helpful to optimize parameter configurations and better design fault-detection protocols for edge computing.

Recent advances in 6th Generation (6G) mobile networks can meet the needs of deeper intelligent communication. Meanwhile, it also brings great challenges to the security and privacy preservation of user data. Federated learning is emerging as a distributed learning method to preserve privacy by enabling users to train machine learning models locally and requiring the users to upload only model parameters instead of sending original data to the server. However, the users with mobile devices may be unwilling to participate in federated learning tasks due to considerable overhead for computation and communication, which degrades performance and hinders the application of federated learning. In this study, an iterative double auction is designed based on resource cooperation scheme to incentivize the mobile devices to contribute their resources in federated learning tasks, where model trainers act as sellers and task requesters as buyers. Access point determines optimal training time cost and payment according to the prices offered by the sellers and the buyers. The goal of our incentivizing resource cooperation scheme is to maximize the total utility of federated learning market under information asymmetry. Numerical results show that the proposed scheme can converge to the optimal solution, and also can significantly improve model accuracy and degrade model loss value.

In order to solve the problems of low success rate and long planning time of mobile robot in complex environment, a mobile robot docking position optimization algorithm is proposed based on environment information preprocessing. Firstly, the workspace of the manipulator is analyzed, and the evaluation criteria of grasping difficulty are obtained. The positions of objects, obstacles and mobile chassis in the environment are simplified as points and projected on the xy plane. According to the evaluation criteria of grasping difficulty, the optimized chassis docking position of the mobile robot is obtained. For the obstacle avoidance problem of the manipulator, the rapidly-exploring random trees (RRT) algorithm is used to realize the robot end, connecting rod and obstacle. Finally, through the simulation and experiments in the motion capture system, it is found that using the parking position optimization algorithm of mobile robot can increase the success rate of grasping objects and the speed of grasping planning .

To achieve secure and reliable block verification, miner nodes of Delegated Proof-of-Stake (DPoS) consensus algorithm can collaborate with nearby light nodes (e.g., smart phones) to verify new block data for secure blockchain-empowered vehicular networks. In order to encourage miners to actively cooperate with light nodes in block verification, a Stackelberg game model is proposed to formulate the interaction between blockchain users and miners, thus jointly maximizing the utility of blockchain users and the profits of miners. The blockchain user acts as the leader setting the optimal transaction fee for block verification, and the miners as the followers determining the optimal number of verifiers to be recruited for block verification. To find out the Nash equilibrium of the game model, a multi-agent reinforcement learning algorithm is designed to search for a strategy close to the optimal one. The numerical results show that the proposed scheme can jointly maximize the benefits of blockchain users and miners and also ensure the safety and reliability of block verification.

Automated guided vehicle (AGV) autonomous path planning is an important part of the logistics system in discrete manufacturing smart factories. AGV can greatly improve the intelligence and automation capabilities of discrete smart manufacturing. The traditional AGV navigation method has a low degree of freedom. The autonomous path planning of AGV is studied under the scenario of discrete manufacturing smart factories, and deep reinforcement learning methods applied to improve the freedom of autonomous path planning. A neural network structure for multi-modal environmental information perception is designed, and the path planning policy of AGV under global obstacles introduced to the path planning in the complex discrete manufacturing smart factory scenario, thereby realizing the AGVs end-to-end path planning from environmental perception to action for decision making. The experimental results show that AGV can independently plan paths in the complex and unknown intelligent logistics system environment of discrete manufacturing smart factories, and has a high success rate and obstacle avoidance ability.

Conventional master/slave-based data processing frameworks are vulnerable to single point of failure and performance bottlenecks of the master node. In contrast, blockchain systems adopt a decentralized framework and are capable of aggregating enormous computing resources. A blockchain-based data processing framework is proposed that utilizes the advantages of the blockchain for solving the drawbacks of the centralized framework. In this framework, the blockchain stores the task information and the adopted proof of useful work consensus enables nodes to process tasks using their computing resources, while competing for the leader (who dispatches pending tasks to the blockchain). Extensive simulations show that the proposed framework is better than the centralized framework in terms of the throughput and the task response time.

To study the way to realize the intrinsic safety of production system, the method of realizing the intrinsic safety of production system by establishing artificial intelligence management system is proposed. Firstly, the concept and problems of intrinsic safety are discussed, including the functions and characteristics of human, machine, environment and management subsystems in the production system, and their obstacles to the realization of intrinsic safety also studied; secondly, the feasibility of realizing intrinsic safety by artificial intelligence is discussed, and the structure of artificial intelligence production system established. Compared with the original structure, the operator disappears, the role of manager changes and increases feedback mechanism and system complexity is decreased. Finally, the way to realize intrinsic safety is discussed, that is to build an artificial intelligence management system, which has the characteristics of double cycle and self-learning. Although the theory of fault pattern recognition and fault knowledge base is not mature, it is feasible to realize the intrinsic safety of production process by establishing artificial intelligence management system.

In order to reduce the fuzziness of using scientific effect in TRIZ, and to accurately and clearly describe the transformation between input and output in the effect, a formal expression method based on Extenics is proposed to study scientific effect. First, the concepts of extension transformation, conductive effect and scientific effect are briefly introduced, and then a formal quantitative research on the relationship between seven important physical effect and extension transformation and conductive effect is carried out, establishing the various objects involved in physical effect Matter-element model, and establishing the correlation between matter-element according to the domain knowledge and correlation rules in Extenics. An active transformation of matter-element in the correlation relationship is implemented based on the knowledge of active transformation and conductive transformation, obtaining the corresponding conductive transformation. Then according to the calculation formula of physical effect and conductive effect in the domain knowledge, the conversion formula of physical effect and conductive effect is obtained. Finally, methods are summarized to establish the general relationship between physical effects and extension transformation and conductive effect. This study describes the main contents of various physical effects accurately and in detail by formal methods, so that the mechanism of various physical effects and the conversion relationship between input and output can be understood. A basis and methods are also provided for solving the contradictions in the field of engineering technology combined with physical effects and extension transformation, and a new idea is provided for finding new scientific effects.

In order to scientifically evaluate and decide the optimal multimodal combination scheme of driver's emotion perception device, based on the extension design multi-functional product creation method, the detection function of emotion perception device is analyzed by using the divergence analysis method of basic element theory, and the matter-element like model of specific product established. The relationship between the three installation positions of a single detection device is analyzed, and a series of reorganization schemes are developed. The optimization evaluation method is used to screen them, and an effective combination scheme of multimodal driver emotion perception device is determined.

Idea generation is a key aspect in design, but it is difficult to obtain systematic ideas due to the limited skills of individuals. The method of studying the treatment of contradictory problems in Extenics can provide a reference for solving the problems in design. Taking Ouzhuang metro station in Guangzhou as an example to solve the problem of littering cigarette butts affecting the public living environment, an Extension innovation method is introduced to model the problem with basic elements, and based on the analysis of the actual situation of Ouzhuang metro station, the problem is studied by detailed expansion analysis on multiple basic elements, and using the extension innovation method such as transformation to obtain more than 13 solutions.Finally, the application in a small area of Ouzhuang metro station proved the effectiveness of the solutions. The method can provide new directions for generating systematic ideas for problem solving in design.