Abstract:In this paper,a class of errors-in-variables(EIV)model identification problems is discussed based on the v-gap measure to obtain the L2 optimum,and it includes EIV model identification for single input and single output(SISO)open-loop and closed-loop systems as well as multiple input and multiple output(MIMO)systems.Basing the method on the v-gap measure involves considering the distance between the disturbed and approximation models as an optimization criterion and then obtaining the optimal solution to the model identification problem when the corresponding Nyquist winding condition is met.Obtaining the L2 optimum involves performing an orthogonal decomposition in the L2 space of the disturbed output and input frequency-domain experimental data such that the range space of the system model,described by the normalized right coprime factors,is orthogonal to that of the noise model,described by the system model's complementary factor.The associated approximation error(noise)can then be indirectly minimized in the L2 sense by optimizing the v-gap measure.The proposed method offers a new approach for EIV model identification and is applicable to both closed-loop and multivariable systems.It requires no assumptions about the characteristics of the bounded system input and output noises and can simultaneously estimate the system and associated noise models,making it broadly applicable in engineering practice.

Abstract:With the increasing prevalence of cooperation among manufacturing enterprises,distributed manufacturing characterized by the optimal sharing of resources has emerged as a modern production paradigm.As a core resource in production manufacturing,efficient scheduling of material supply between suppliers and manufacturers can significantly enhance distributed production efficiency and reduce production costs.To addresse the distributed production material supply scheduling problem,a data-model driven intelligent optimization approach is proposed to simultaneously optimize both manufacturer satisfaction and material tardiness.Firstly,a mixed-integer programming model incorporating a dy-namic replenishment mechanism is formulated for the complex supply network comprising multiple warehouses,factories,and material types.Secondly,the mathematical solver Gurobi and heuristic rules are employed respectively to maximize satisfaction and minimize tardiness for the small-scaled problems and large-scaled problems.Thus,two high-quality single-objective optimal solutions are yielded as the start and end points for multi-objective optimization.Thirdly,an adaptive path-relinking mechanism is designed based on initial solutions,utilizing a difference-driven adaptive exploration strategy to enhance diversity of the multi-objective solutions.Finally,a goal-driven local intensification is proposed to further im-prove exploitation.Experimental results on the instances with varying scales demonstrate that the proposed algorithm can effectively solve the distributed material supply scheduling problem.

Abstract:Distributed optimization,as a critical research topic in the field of multi-agent cooperative control,focuses on achieving global collaborative optimization objectives through localized information exchange among agents.Under distributed optimization,each agent utilizes its own local information and communicates with neighboring nodes via a communication network to collaboratively minimize the global objective function.This study conducts a systematic review of the relevant literature from recent years and provides an overview of the current research status in both offline and online distributed optimization.Finally,we discuss several future research directions and highlight the potential developments in distributed optimization for multiagent systems.

Abstract:In the era of Industry 5.0,human-centred intelligent manufacturing has become a hot research topic.For the dual-resource flexible job shop scheduling problem,studies have considered the learning effect of workers,but have not yet addressed the impact of the dual-layer learning effect of workers on jobs and machines on production efficiency.For this reason,this paper proposes for the first time a two-resource flexible job shop scheduling problem considering the dual-layer learning effect of workers,and constructs a mathematical model with maximum completion time as the optimisation objective.To solve the problem,an improved Memetic algorithm is proposed,and the main improvements include:De-signing a three-layer encoding that meets the problem characteristics,proposing an active scheduling decoding strategy to improve the solution quality,developing a population initialisation strategy to enhance the diversity,and designing a fusion of cross-variance updating and variable-neighbourhood searching strategies to improve the ability of global exploration and local optimisation.Finally,the effectiveness of the algorithm and model is verified by comparison experiments.

Abstract:In federated learning architectures for mobile crowdsensing,users face the risk of privacy leakage.Existing differential privacy-based schemes suffer from a loss of local model training accuracy due to gradient clipping,especially in heterogeneous environments.To address these issues,our paper first employs federated stochastic principal component analysis to reduce the dimensionality of the data.Subsequently,the objective function perturbed by rényi-differential privacy is used to replace the gradient for updates.Then,Bregman divergence is introduced as a regularization term to update the loss function,constraining the deviation between the local and global models.Experimental results demonstrate that the proposed method achieves higher accuracy and convergence precision compared to several existing approaches.

Abstract:In the modal analysis and control of nonlinear dynamical systems,participation factors(PFs)of state variables with respect to a critical or selected mode serve as a pivotal tool for simplifying stability studies by focusing on a subset of highly influential state variables.For linear systems,PFs are uniquely determined by the mode's composition and shape,which are defined by the system's left and right eigenvectors,respectively.However,the uniqueness of other types of PFs has not been thoroughly addressed in literatures.This paper establishes sufficient conditions for the uniqueness of nonlinear PFs and five other PF variants,taking into account uncertain scaling factors in a mode's shape and composition.These scaling factors arise from variations in the choice of physical units,which depend on the value ranges of real-world state variables.Understanding these sufficient conditions is essential for the correct application of PFs in practical stability analysis and control design.

Abstract:The insufficient accuracy of energy consumption modeling in the copper foil electrolysis process compromis-es the reliability of energy-saving optimization,posing significant challenges to existing optimization methods.To address this issue,particularly given the difficulty of further enhancing modeling accuracy,this study proposes a novel energy-saving optimization approach for copper foil electrolysis based on local multi-scale filtering and systematically analyzes its reliability.Firstly,to investigate the impact of modeling accuracy on energy consumption optimization,models with varying levels of accuracy are generated through a loop-judgment structure,which enables analysis of the underlying influ-ence mechanisms.Secondly,a reliability analysis method is developed based on the goodness-of-fit metrics from existing regression models,enabling a quantitative assessment of the reliability of energy-saving optimization.Additionally,an improved strategy employing local multi-scale filtering is introduced to mitigate the influence of prediction errors on the intelligent bionic optimization algorithm,thereby addressing reliability issues arising from modeling accuracy limitations;relevant theoretical derivations are also provided.Finally,an industrial experiment in a copper foil electrolysis enterprise demonstrates that the proposed method reduces the absolute error of energy consumption optimization from 4.29％to 1.53％,indicating that this approach can substantially enhance the reliability of intelligent bionic optimization algorithms even with lower model accuracy.

Abstract:Existing learning-based constructive task allocation methods require continuously generating a complete task allocation scheme before assigning tasks to agents,which fails to meet the real-time demands of large-scale urgent scenarios such as rescue or confrontation.To address this,a multi-agent batch task allocation method based on deep reinforcement learning is proposed in this paper.In this method,a policy model including an encoder,agent and task-node selection decoders,and a recursive embedding structure is designed that can generate a batch of partial task allocation schemes constructed by agent-task node pairs simultaneously according to the objective function's optimality requirements.In online task allocation,agents no longer need to wait for the complete task allocation scheme before executing the tasks.The evaluation results showed that the proposed method improves the real-time performance,reliability,and cooperative capability of task allocation in urgent scenarios.

Abstract:Active disturbance rejection control(ADRC)brings new concepts such as internal disturbance,total dis-turbance,and extended state,while interpreting these concepts becomes a new problem of control theory.Taking the second-order plant without zero as an example,this paper interprets the above concepts in the framework of linear system theory.Internal disturbance is interpreted as the signal generated by the feedback item in the controllable canonical form.The concept of Han extended realization is proposed,while extended state,which has a state variable observable but un-controllable,is the state of Han extended realization.Furthermore,extended state observer is the observer of Han extended realization,while linear ADRC is output feedback based on Han extended realization.These results show that the above concepts are totally consistent with linear systems theory,and the research of linear ADRC may bring new directions and vitality into linear system theory.

Abstract:The intraday dispatch optimization model is a nonlinear multi-objective problem involving semi-continuous variables,while also imposing high requirements on computational efficiency.This poses a challenge to the convergence speed of existing multi-objective intelligent optimization algorithms.To address this,this paper constructs an intraday dispatch model with two time granularities(15-minute and 5-minute intervals)and a corresponding optimization frame-work.Given the high timeliness requirement of intraday dispatch optimization,a self-adaptive update strategy for the fully connected weight network model is designed based on the two-fully-connected-weight-network evolutionary algorithm(TFCWNEA),denoted as S-TFCWNEA.By adaptively adjusting the standard deviation of model parameters during the optimization process,the strategy balances global and local search capabilities,accelerating population convergence.The two-time-granularity dispatch optimization achieves a stepwise refinement of the day-ahead scheduling plan.Simulation results demonstrate that the proposed self-adaptive update strategy for the fully connected weight network significantly improves the algorithm's convergence speed,enabling rapid response to fluctuations in renewable generation and load demand.The method efficiently adjusts the dispatch plan within a limited time frame.

Abstract:The continuous advancement of aerospace technology has enabled agile earth observation satellites with advanced attitude maneuverability to play a significant role in climate monitoring,military intelligence,and other fields.In order to meet the complex agile satellite task scheduling requirements,this paper investigates a time-dependent multi-objective scheduling optimization problem for multiple agile satellites,aiming to minimize task failure rates and satellite load balancing.Firstly,a mathematical programming model is constructed based on the problem characteristics.Secondly,a hybrid evolutionary game scheduling algorithm(HEGSA)is proposed based on evolutionary game theory,which includes two evolutionary stages:global exploitation and local exploration.In the global exploitation stage,heuristic strategies are employed to generate two subpopulations with heterogeneous identities,and a multi-objective evolutionary game strategy is used to optimize each subpopulation to balance convergence and diversity.In the local exploration stage,a self-learning operator is used to enhance the efficient search of the solution space.Finally,the effectiveness of HEGSA is verified through simulation experiments.

Abstract:Cranes are key heavy-duty material handling equipment widely used in shops,warehouses,ports,and other industrial settings.The scheduling of cranes significantly affects transportation efficiency and the achievement of produc-tion goals.To address the crane scheduling problem with time windows(CSP-TW),a mixed-integer linear programming model based on spatio-temporal discretization is developed.Based on the characteristics of the model,a hierarchical rein-forcement learning(HRL)decision-making framework is designed.The high-level decision network assigns transportation tasks to appropriate cranes,while the low-level network plans paths for each crane to complete its assigned task.During the learning process,action tabu rules are introduced to avoid ineffective actions and guide the decision networks toward the dominant policy space.Subsequently,external experience pooling and the dueling double deep Q-network strategy are adopted to train the decision networks.Tests were executed based on the logistics simulation platform of a steel plant from a certain company.Ablation experiments show that the introduction of action tabu rules improves learning efficiency.Training comparisons indicate that HRL achieves better convergence than the end-to-end framework.Compar-ative experiments demonstrate that HRL outperforms several methods,including multi-rule combinations,meta-heuristic algorithms,end-to-end and deep Q-network,while satisfying second-level response-time requirements for applications.

Abstract:Unmanned aerial vehicles(UAVs),with their air mobility and autonomy,have shown significant value in logistics and distribution.This model can significantly reduce manpower costs and improve the flexibility and response efficiency of the logistics network.However,obstacles such as buildings and mountains in three-dimensional complex environments pose serious challenges to UAV flight safety.How to construct the distribution path under the constraint of obstacle avoidance has become a key issue in the optimisation of UAV logistics system.For the demand of 3D path planning,this study establishes a comprehensive analysis framework containing environment modelling and mathematical modelling,and proposes an improved artificial bee colony algorithm incorporating a reinforcement learning mechanism.The algorithm adopts a heuristic rule based on the spatial relationship between the starting and finishing points to generate the initial population,and dynamically selects three search strategies in the honey bee stage through reinforcement learning,which significantly improves the quality of the initial solution and the directionality of the search.In the observation bee stage,a reverse learning mechanism is introduced to generate complementary populations to enhance the convergence accuracy and speed of the algorithm.Simulation experiments show that compared with the traditional algorithm,the improved algorithm has significant advantages in terms of path cost and computational efficiency,and can provide an efficient solution for UAV logistics path planning in complex 3D scenes.

Abstract:This paper addresses the multiple weeding robots task allocation(MWRTA)problem,aiming to minimize the maximum task completion time,total energy consumption,and the amount of residual pesticides,which are key perfor-mance indicators in sustainable agricultural systems.A mixed integer linear programming(MILP)formulation is proposed,and a novel grouping strategy-based multi-objective discrete artificial bee colony algorithm(GMO-DABC)is developed for solving the MWRTA problem efficiently.Firstly,heuristic methods integrating grouping strategy with load balancing is designed to effectively generate solutions.Secondly,neighborhood operators are designed based on the grouping strategy,dynamically adjusting neighborhood structures by knowledge-guided to reduce the risk of local optimum.Finally,a search strategy combining grouping strategy with non-dominated frontier analysis is proposed to efficiently explore the solution space.Extensive simulation experiments under multiple instance scales validate the superiority of GMO-DABC over several state-of-the-art algorithms in terms of solution quality,convergence speed and robustness,confirms its strong optimization capability and practical value for real-world agricultural applications.

Abstract:Aiming at the nonlinear,multivariable,strongly coupled and multi-source interference dynamics character-istics of tilt-rotor electric vertical take-off and landing vehicle(eVTOL),this paper proposes a fast terminal sliding mode control algorithm based on fast convergence law.Firstly,based on the established six-degree-of-freedom nonlinear model,the actual control quantities are transformed into pseudo-control variables for each control channel,and the flight control is decoupled through the pseudo-control variables.Secondly,the fast terminal sliding mode surface is constructed,the equivalent control part in the controller is designed to achieve the fast convergence of the system state tracking error in a finite period of time,and a new type of successive convergence law is designed based on the fal nonlinear function to further reduce the chatter.Then an exponentially convergent disturbance observer is designed to compensate for the pos-sible disturbances of the system.Finally,the simulation results show that the proposed control algorithm shows superior trajectory tracking and anti-disturbance capabilities in the face of complex disturbances.

Abstract:To address the practical finite-time bipartite consensus problem of general linear multi-agent systems where the agents' state information is not fully accessible and communication resources are limited,a Luenberger observer is introduced by adopting output feedback approach,and a distributed dynamic event-triggered finite-time bipartite consensus protocol is designed based on the estimated observer state.This protocol takes into account the coexistence of cooperative and competitive relationships in the network topology,effectively overcoming the limitations of agent state measurement in practical applications and significantly reducing the communication frequency of the system.Meanwhile,utilizing finite-time stability theory,algebraic graph theory,and matrix theory,sufficient conditions are derived under which the multi-agent system can achieve practical finite-time bipartite consensus with the designed control protocol,and it is proven that the system is free from Zeno behavior.By constructing a dynamic event-triggered mechanism that matches the finite-time bi-partite consensus control,the proposed control algorithm exhibits satisfactory advantages in saving system communication resources.Finally,the feasibility and effectiveness of the theoretical results are verified through simulation examples.

Abstract:The application of permanent magnet linear synchronous motor(PMLSM)in periodic high-precision po-sition servo is susceptible to uncertainties,which is primarily caused by periodic end effects thrust ripple and parameter perturbations.Therefore,a method for H∞ preview repetitive control of PMLSM based on an equivalent input disturbance compensation is proposed.Firstly,a preview compensator and an error compensation term are introduced into the feed-forward compensation loop,which are used to minimize the tracking error caused by the ineffectiveness of the positive feedback delay link in the first control cycle of the repetitive controller.Then,an equivalent input disturbance compensator equivalent input disturbance compensator based on the sliding mode observer(SMO-RCEID)is designed based on the sliding mode observer to compensate for the end effects thrust ripple and parameter perturbations of the PMLSM.Finally,the conditions for the stability of the SMO-RCEID compensator and the H∞ preview repetitive control system described by the two-dimensional continuous/discrete hybrid model are derived by means of the Lyapunov stability theory combined with the linear matrix inequality(LMI),and the controller gain and observer gain of the system are obtained.Simulation results demonstrate that this method has excellent tracking performance and disturbance rejection capability.

Abstract:Along with the implementation of the"double carbon"goal,the proportion of new energy power generation units in the power system has gradually increased.Distributed smart grids achieve coordinated operation between source,grid,load,storage,and charging by integrating intelligent sensing,communication,decision-making,and control tech-niques.This will form a more efficient,stable,and reliable power system.Under the new power system,economic dispatch problems show strong distributed characteristics.Distributed economic dispatch seeks to balance supply and demand con-straints,and distribute power by facilitating information exchange among power generation units.It focuses on achieving a cost-efficient allocation of power generation under specified constraints.This paper employs a precise first-order con-sensus tracking method with momentum acceleration to quickly obtain dual variables for the equality constraints.It then utilizes projection operators for mapping these dual variables to power allocation to attain the lowest power generation cost-s while adhering to system coupling constraints and local inequality restrictions.Convex optimization theory and matrix contraction mapping are adopted to demonstrate the convergence of the proposed algorithm.Unlike existing methods,our developed approach requires each agent to exchange only a single dual variable with its neighboring agents,and moreover,the choice of control parameter depends solely on the strongly convex coefficient.Two simulation scenarios are presented to validate the efficiency and convergence rate of the algorithm developed in this study.

Abstract:Up to now,the fixed-time synchronization analysis of inertial neural networks has mainly utilized variable substitution method,which is effective,but expands the system's dimension and eliminates the influence of the inertial term.In order to consider the intuitive effects of the inertia term,this paper discusses the fixed-time synchronization of delayed fuzzy neural networks with inertia terms via using the non-reduced-order approach.Under the Filippov solution framework and finite time stability theory,some synchronization criteria are obtained to ensure the realization of fixed-time synchronization of the proposed neural system by designing nonlinear feedback controller.In addition,the upper bound of the synchronization time is estimated by using some inequality techniques,which can provide reliability guarantee for its application in practical engineering.Finally,the results obtained in this article are verified through numerical examples and an image encryption application.

Abstract:For the electric two-drive differential unmanned vehicle,due to the existence of road friction,vehicle center of gravity instability and other factors,the driving resistance of the left and right driving wheels is different,resulting in the driving direction of the vehicle always deviates from the longitudinal axis of the unmanned vehicle regularly to the left or right.The traditional driving scheme of unmanned vehicles relies on images,base station positioning and other methods to obtain the offset error according to the external reference environment and then adjust the body.Such methods have limited ability to cope with the complex environment.This paper presents a hierarchical control strategy for unmanned vehicles based on active disturbance rejection control,which relies only on wheeled odometer.Firstly,the offset distance is calculated according to the odometer information of the differential unmanned vehicle,and the upper controller designs the expected course angle function according to the offset error,and adjusts the course angle of the vehicle to achieve the purpose of driving control.The lower controller designs an extended state observer based on the dynamics model to estimate the disturbance caused by vehicle motion and compensate the motors on both sides.Finally,the effectiveness of the proposed hierarchical control strategy is verified by numerical simulation and real vehicle test.