Over 50 million people globally suffer from Alzheimer's disease (AD), emphasizing the need for efficient, early diagnostic tools. Traditional methods like Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) scans are expensive, bulky, and slow. Microwave-based techniques offer a cost-effective, non-invasive, and portable solution, diverging from conventional neuroimaging practices. This article introduces a deep learning approach for monitoring AD , using realistic numerical brain phantoms to simulate scattered signals via the CST Studio Suite. The obtained data is preprocessed using normalization, standardization, and outlier removal to ensure data integrity. Furthermore, we propose a novel data augmentation technique to enrich the dataset across various AD stages. Our deep learning approach combines Recursive Feature Elimination (RFE) with Principal Component Analysis (PCA) and Autoencoders (AE) for optimal feature selection. Convolution Neural Network (CNN) is combined with Gated Recurrent Unit (GRU), Bidirectional Long Short Term Memory (Bidirectional-LSTM), and Long Short-Term Memory (LSTM) to improve classification performance. The integration of RFE-PCA-AE significantly elevates performance, with the CNN+GRU model achieving an 87% accuracy rate, thus outperforming existing studies.

In a two-crane cooperative lift process, unreasonable load distribution on the two cranes may cause one of the cranes to overload, which may cause a dangerous overturn accident. Therefore, the load distribution should be taken as a constraint to yield a safe path for a cooperative lift. Besides, the load distribution on the two cranes varies with the changing postures of the cranes. However, the explicit relationship between the load distribution and the postures has not been reported. Therefore, this article first presents a relationship model between the postures of the two cranes and the load distribution on them. Next, a new path-planning method based on reinforcement learning is explained, which utilizes the load constraint as the optimization object in the cooperative two-crane lift. Simulation results show that the new method yields a short lift path with reasonable load distribution.

This article investigates optimal consensus problems of switching multiagent systems (MASs). For such kind of MASs, local neighborhood tracking error (LNTE) system is time varying because of the switching communication graph (CG). Existing performance index defined on the LNTE system is thus invalid for the switching MASs. This article addresses this problem by defining a new augmented LNTE system. The augmented LNTE system is constructed using the joint CG and is thus time-invariant. Subsequently, the optimal consensus problems for the MASs are formulated using the augmented LNTE system. Value iteration algorithm that employs an actor-critic network is used to learn the optimal controller. The article provides a theoretical analysis demonstrating the learning stability and control stability of the value iteration method. Furthermore, we also show that the MASs will reach approximate Nash equilibrium. Simulation results proves the effectiveness of the proposed method.

In the geological drilling process, predicting the rate of penetration (ROP) is significantly important for improving drilling efficiency and reducing nondrilling time. However, due to the drilling data pollution and the complex nonlinearity in the geological drilling process, a reliable and highly accurate ROP prediction model is not easy to construct, and the model accuracy is affected by the value of model hyperparameters. In order to overcome the difficulties in modeling, a novel ROP model is developed to handle abnormal data and achieve nonlinear fitting. First, a local outlier factor is introduced to automatically detect the abnormal data, and then, replace it with the nearest normal data. Then, the support vector regression (SVR) method is applied to construct nonlinear prediction model for ROP, and a modified bat algorithm (MBA) is developed to solve the non-convex problem in determining optimal value of hyperparameters for SVR-based ROP model. The MBA has six modifications to improve the global search ability, which achieves better performance in global search ability compared with other nine algorithms based on the experiments of IEEE 2005 benchmark functions. The developed ROP prediction model that combines SVR and MBA methods is tested based on actual drilling data and a semiphysical system, and the simulation and application results show that the developed modeling method has higher ROP prediction accuracy compared with the other modeling methods.

This article investigates the optimal synchronization problem for unknown discrete-time nonlinear heterogeneous multiagent systems (MASs). It is very intractable to derive the analytical solutions of coupled Bellman's equations, which are necessary to overcome this problem. We propose a data-based optimal synchronization control strategy based on a hierarchical and distributed optimal control framework composed of a model reference adaptive control (MRAC) layer and a distributed control layer. In the MRAC layer, the similar-offline MRAC algorithm is developed to make subsystems of MASs track their reference models, respectively. Then, the distributed optimal control problem of nonlinear heterogeneous MASs is transformed into that of homogeneous MASs composed of the reference models and the leader. In the distributed control layer, the distributed reference policy iteration algorithm is proposed to derive the solutions of coupled composite nonlinear Bellman's equations, which ensure that the homogeneous MASs reach synchronization with optimum. The suboptimal synchronization control is achieved via optimization further. Convergence analysis of both algorithms is rigorously provided. The simulation results verify the effectiveness of the proposed strategy.

This article is concerned with the optimal synchronization problem for discrete-time nonlinear heterogeneous multiagent systems (MASs) with an active leader. To overcome the difficulty in the derivation of the optimal control protocols for these systems, we develop an observer-based adaptive synchronization control approach, including the designs of a distributed observer and a distributed model reference adaptive controller with no prior knowledge of all agents' dynamics. To begin with, for the purpose of estimating the state of a nonlinear active leader for each follower, an adaptive neural network distributed observer is designed. Such an observer serves as a reference model in the distributed model reference adaptive control (MRAC). Then, a reinforcement learning-based distributed MRAC algorithm is presented to make every follower track its corresponding reference model on behavior in real time. In this algorithm, a distributed actor-critic network is employed to approximate the optimal distributed control protocols and the cost function. Through convergence analysis, the overall observer estimation error, the model reference tracking error, and the weight estimation errors are proved to be uniformly ultimately bounded. The developed approach further achieves the synchronization by means of synthesizing these results. The effectiveness of the developed approach is verified through a numerical example.

Drilling trajectory optimization (DTO) is an important step in intelligent control of directional drilling process. DTO in complex multiple strata geological environment is a new nonlinear, strongly constrained, black-box multi-objective optimization problem with discrete objectives. In this paper, a multi-objective evolutionary algorithm (MOEA) optimization framework is present for an obstacle avoidance drilling trajectory design model. First, a new three-segment trajectory model is established by natural curve method to measure the whole trajectory positions. Meanwhile, a drilling point wellbore stability model is derived by Mohr-Coulomb failure criteria. Based on this calculation model, the trajectory data is mapping to get a multi-strata parameter field. Several representative MOEAs are adopted to design the highest wellbore stability trajectory in this multi-strata geological environment. A case study from actual drilling site shows that the result of adaptive non-dominated sorting genetic algorithm (ANSGAIII) is better than that of traditional trail and error drilling trajectory design, which has a good application prospect in drilling engineering.

This paper investigates the consensus problems of heterogeneous discrete-time (DT) nonlinear multi-agent systems (MASs) with unknown dynamics and switching typology. By adding a virtual model to each agent and using the model reference protocol to the actual model, the consensus problems of unknown nonlinear MASs are transformed into consensus problems of known virtual linear MASs, and a distributed control law is designed for the virtual model to make the virtual linear MASs achieve consensus under the switching typology. Two numerical simulations with nonidentical nonlinear dynamics and switching typology are given to prove the effectiveness of the proposed method.

Wind is the dominant factor for wind power generation. However, wind, which is intermittent and fluctuating all the time, is hard to be accurately forecasted, especially only by single-source numerical weather prediction. This article presents a short-term wind power prediction method based on multisource wind speed fusion. First, the relationships among the weather factors and wind power, and the characteristics of three independent forecasted wind speed (FWS) provided by three organizations are analyzed. Next, a weighted naive Bayes method is described to fuse the three FWSs in order to estimate an accurate wind speed according to their characteristics. Then, a backpropagation neural network is designed to predict the wind power based on the fused wind speed. Finally, application results show that the method accurately predicts wind power generation, and the accuracy is much higher than that predicted by conventional methods.

To reduce energy consumption and harmful emission, it is of great significance to improve carbon efficiency in sintering process, which is able to be achieved if the carbon efficiency can be accurately predicted. In this article, the ratio of CO and CO2(CO/CO2) is taken as a measurement of the carbon efficiency. As CO/CO2is hard to measure, and there exist multiple working conditions, multiple variables, and nonlinearity, a hybrid CO/CO2prediction model is devised based on the aforementioned characteristics. First, the sintering process is analyzed, and the key characteristics to predict the CO/CO2are extracted. Next, the configuration of the prediction model is given based on the analysis. The model consists by two submodels, one is to predict the state variables by an improved just-in-time learning model, combining three neural network (NN) models. The other is to predict CO/CO2with semisupervised algorithm, based on deep belief network with a combination of the three NN regression methods. Then, the configurations of the two submodels are introduced in detail. The test results based on actual running data exhibit the good performance of the model.

This article concerns the distributed optimal observer design, which aims to estimate the discrete-time nonlinear leader state for all followers with optimum online. The communication constraint and no prior knowledge of the leader’s dynamics cause a challenging task about the observer design. To this end, an adaptive critic design-based distributed optimal observer is developed via the actor–critic framework. The critic network is employed to approximate the cost function. The action network produces the optimal correction policy of the observer. Through convergence analysis, the overall estimation error and the weight estimation errors of the critic network and the action network are demonstrated to be all uniformly ultimately bounded (UUB). The simulation results verify the effectiveness of the developed observer.

This paper focuses on multi-agent systems under the disturbances of parameter uncertainties, noises and external disturbances simultaneously, which are more suitable for the changing environments than most ideal models. For this case, the robust H∞ consensus control problem is studied to attenuate the influence of disturbances below a given level. We aim to design a consensus controller such that the closed-loop muti-agent systems reach the desired H∞performance. By tools of Kronecker product, graph theory, and infinite horizon bounded real lemma, sufficient conditions for existence of H∞consensus protocol design are obtained in terms of linear matrix inequalities (LMIs). Finally, a numerical example is given to show the validity of the method.

Internet of Vehicles (IoV) is a typical application of Internet-of-Things (IoT) technology in the field of intelligent transportation systems. In the actual IoV, such as the autonomous vehicle fleet, there exists the problem of heterogeneous multivehicle coordination based on IoT communication. How to ensure the synchronization of multiple vehicles is a hot issue. In particular, when the system can only obtain a partial state of the vehicle, and does not know the dynamic model, including the vehicle itself and the companion model. To overcome these deficiencies, this article deals with the model-free output consensus control problem for a class of partially observable heterogeneous multivehicle systems (MVSs). Using measurable input/output data without any system knowledge, this article develops a Q-function-based adaptive dynamic programming (ADP). First, an adaptive distributed observer is designed to estimate the output of the leader. The augmented state representation is built using historical measurable input/output data instead of the unmeasurable inner system state. Then, a Q-function-based ADP method using measurable input/output data was introduced. The method is used to solve this distributed tracking control problem without the requirement for the MVSs dynamics. The convergence analysis of the proposed method is also given. To facilitate the implementation of the proposed method, an actor-critic framework is adopted to approximate the optimal Q-functions and the optimal control policies. It shows that the approximated control policies achieve the distributed optimal tracking control. Finally, the simulation results verify the effectiveness of the developed method for solving multivehicle formation control problems.

Conventionally, as the system's dynamics is known, the optimal consensus control problem relies on solving the coupled Hamilton-Jacobi-Bellman (HJB) equations. In this paper, with the unknown system dynamics being considered, a local Q-function-based adaptive dynamic programming method is put forward to deal with the optimal consensus control problem for unknown discrete-time nonlinear multiagent systems by approximating the solutions of the coupled HJB equations. First, a local Q-function is defined, which considers the local consensus error and the actions of the agent and its neighbors. Using the Q-function, it is convenient to get the derivatives with regard to the weights of the consensus control policies, even without the model of system dynamics. Then, with the defined local Q-function, a distributed policy iteration technique is developed, which is theoretically proved to be convergent to the solutions of the coupled HJB equations. An actor-critic neural network framework for implementing the developed model-free optimal consensus control method is constructed to approximate the local Q-functions and the control policies. Finally, the feasibility and effectiveness of the developed method are verified by a series of simulations.

Adaptive dynamic programming (ADP) is a prevalent way to solve the coupled Hamilton-Jacobi-Bellman (HJB) equations of the optimal consensus control for multi-agent systems (MAS). Neural networks (NNs) are normally used to approximate the value functions in ADP. However, NNs with manually designed features may influence the approximation ability. In this study, kernel-based methods which do not need to set the value function model structure in advance are adopted for value functions approximation. Moreover, to overcome the deficiency that most of the system dynamics are unknown, or the system is too complex to obtain the accurate dynamics. Local action value functions are defined, and kernel-based methods are used to approximate the local action value functions. Thus, an action dependent heuristic dynamic programming (ADHDP) approach using kernel-based local action value functions approximation is developed to achieve the optimal consensus control model-freely. The developed approach uses historical sample data to learn the system dynamics, and avoids the traditional system identification scheme. Simulation results are provided to demonstrate the effectiveness of the presented approach.

SUMMARY The curse of dimensionality is a ubiquitous problem for multiagent reinforcement learning, which means the learning and storing space grows exponentially with the number of agents and hinders the application of multiagent reinforcement learning. To relieve this problem, we propose a new framework named as optimal tracking agent (OTA). The OTA views the other agents as part of the environment and uses a reduced form to learn the optimal decision. Although merging other agents into the environment may reduce the dimension of action space, the environment characterized by such form is dynamic and does not satisfy the convergence of reinforcement learning (RL). Thus, we develop an estimator to track the dynamics of the environment. The estimator obtains the dynamic model, and then the model‐based RL can be used to react to the dynamic environment optimally. Because the Q‐function in OTA is also a dynamic process because of other agents’ dynamics, different from traditional RL, in which the learning is a stationary process and the usual action selection mechanisms just suit to such stationary process, we improve the greedy action selection mechanism to adapt to such dynamics. Thus, the OTA will have convergence. An experiment illustrates the validity and efficiency of the OTA.Copyright © 2012 John Wiley & Sons, Ltd.

To cope with the curse of dimensionality, an ubiquitous problem in multi-agent reinforcement learning, this paper deals with the multi-agent learning in a new perspective and proposes a new algorithm, the optimal tracking agent (OTA). The OTA treats the other agents as a part of the system and uses an estimator to track the dynamics of the system. Thus, it obtains the dynamic model with limit accuracy and uses the model-based reinforcement learning to react optimally to the system. All the processes are just from one agent's perspective, then the searching space for action is just its own and not exponential with the number of agents any more. Thus, the curse of dimensionality is relieved from action space. Experiment illustrates the validity and efficiency of the proposed method.

Particle swarm optimization (PSO) is a population-based stochastic recursion procedure, which simulates the social behavior of a swarm of ants or a school of fish. Based upon the general representation of individual particles, this paper introduces a decreasing coefficient to the updating principle, so that PSO can be viewed as a regular stochastic approximation algorithm. To improve exploration ability, a random velocity is added to the velocity updating in order to balance exploration behavior and convergence rate with respect to different optimization problems. To emphasize the role of this additional velocity, the modified PSO paradigm is named PSO with controllable random exploration velocity (PSO-CREV). Its convergence is proved using Lyapunov theory on stochastic process. From the proof, some properties brought by the stochastic components are obtained such as "divergence before convergence" and "controllable exploration." Finally, a series of benchmarks is proposed to verify the feasibility of PSO-CREV.

It is a challenging task for multiple robots working together to realize object transportation. This paper studies a practical situation that a group of mobile manipulators are used to transport an object whose mass can not be ignored. From the viewpoint of formation, a leader-follower type control is designed. To overcome parameter uncertainty in modeling robot, a decentralized control law is applied to individual robots, in which an adaptive NN is used to model robot dynamics online. Using Lyapunov theory, we have proved that if all end-effectors of robots will keep proper relative distances with a regular formation to manipulate a heavy object, the object will be transported to the destination at last. But due to parameter uncertainties, there may exist a static position error, which can be reduced by proper selection of control coefficients.

The formation task achieved by multiple robots is a tough issue because of the limitations of the sensing abilities and communicating functions among them. Due to an individual robot can only handle local information, an adjacency graph is applied to describe the relationship among multiple robots. Since the relative positions among robots change from time to time, the topology graph describing information exchange is variant. A local control strategy is proposed for an individual robot based on NN control with robust terms. It has been proved that under an assumption of adjacency matrices associated with interaction graph being always connected, the system will converge based on the individual control strategy and multiple robots can construct an unique formation even if interaction topology is variant.