This paper focuses on solving the simultaneous data quantization and data dropouts problems in multi-input multi-output unknown nonlinear discrete systems, with an emphasis on their full tracking control. First, an innovative quantization model-free adaptive iterative learning control scheme is proposed based on the dynamic linearization model of the system. Using a uniform quantizer with an improved encoding-decoding mechanism to quantise the system output, modelling the data dropouts problem as a Bernoulli sequence, and employing intermittent updating to deal with the effects of the random data dropouts problem in the system output, the inherent errors due to data quantization and data dropouts are successfully eliminated, and it is verified that only a small amount of input/output data is required to achieve the tracking and controlling error convergence. Finally, the effectiveness of the method can be demonstrated through theoretical derivation and simulation experiments.
This paper considers the problem of target search in unknown environments and proposes a Unmanned Aerial Vehicle (UAV) cooperative target search based on Behavior Expression Tree (BET). Using behavior expression tree as the structure of UAV motion control strategies helps handle unknown and complex environments, effectively solving problems such as target search, group response and obstacle avoidance. A Heuristic Optimization Algorithm based on Behavior Expression Tree (HOA-BET) is proposed to train and generate an effective behavior expression tree. Multiple tree update operators are designed to help the tree population update and evolve. The relationship between target search efficiency and UAV trajectory rationality is balanced through a mixed approach of multiple indicators. The effectiveness and reliability of the algorithm and control strategy have been demonstrated through comparison with various methods. Experiments in various scenarios have shown the adaptability of the control strategy generated by the algorithm to the environment. Additionally, it has good results when applied to large-scale target search.
In recent works, direct data-driven approaches based on Willems' Fundamental Lemma have become popular in control theory. Compared to static output feedback, the approach using past finite-length input-output trajectories to construct feedback controller has a better global convergence guarantee for linear quadratic regulator problem. This paper proposes a direct data-driven policy gradient method, which can be applied to design an optimal output feedback controller. We begin by applying a non-minimum state representation composed of past input-output data to data-parameterize the original problem. Based on this, we present a novel data-driven policy optimization method to directly update the policy, where the gradient is explicitly computed by using persistently exciting raw data. Moreover, we prove that this data-driven policy gradient method exhibits global sublinear convergence. The paper concludes by presenting simulations that validate our theoretical results.
In this paper, the recursive state estimation problem is discussed for a class of nonlinear complex networks with random coupling strength (RCS) and random sensor saturation (RSS) under weighted try-once-discard protocols. The coupling strength between nodes follows a uniform distribution, and the random saturation phenomenon of sensors is described by a set of Bernoulli distribution variebles. The main purpose of this paper is to design an estimator that guarantees that the covariance of the estimation error has an upper bound when the RCS and RSS exist simultaneously, and this upper bound is minimized at each time instant. Finally, a simulation example is provided to prove the effectiveness of the proposed recursive state estimation scheme.
In this paper, we discuss the closed-loop Stackelberg strategies for linear-quadratic leader-follower games. Different from previous research, our approach adopts a linear closed-loop form with a two-step memory. The primary challenge arises from the leader's encounter with a nonclassical control problem involving constraints. To overcome the difficulty, we first attribute the solvability of the leader-follower games to the forward and backward difference equations by applying the constrained maximum principle. Subsequently, we present explicit linear closed-loop Stackelberg strategies with a two-step memory, based on coupled Riccati equations, achieved by decoupling the forward and backward difference equations. The key technique lies in employing quadratic optimization to determine the gain matrix of the linear closed-loop Stackelberg strategies. Numerical examples substantiate the superiority of our proposed strategies over feedback and one-step memory strategies.
In order to enable users to conveniently query medical knowledge online and promote the construction of smart healthcare, a medical Chinese named entity recognition model based on RoBERTa-BiSRU-CRF is designed. Firstly, the model uses RoBERTa to extract feature word vectors of medical data; secondly, BiSRU model is used to extract high-dimensional global sequence features of medical text; finally, conditional random field is used to output the global optimal label sequence. Finally, the tag sequence results are sent to the Medical Knowledge Graph to query the answers. Validated by experiments on the CMeEE medical dataset. The experimental results show that the proposed RoBERTa-BiSRU-CRF entity recognition model achieves an accuracy of 95.16%, which is significantly better than the other models in the experiments, proving the effectiveness of the model.
We study the contraction analysis-a notion that ensures the system's trajectories converge exponentially to a certain trajectory-of discrete-time singular nonlinear switched systems under a fully known switching signal. A necessary and sufficient condition for the system to be contracting is provided via a single Lyapunov function approach, which is rather conservative due to the requirement of the existence of a common Lyapunov function for all subsystems. For a less conservative condition, a sufficient condition is then provided via a switched Lyapunov function approach. This approach does not require a common Lyapunov function for all subsystems, nevertheless, it requires that all subsystems are contracting. Finally, the study is complemented with illustrative examples of contracting switched systems.
As one of the most crucial secondary energy resources in steel industry, Linz Donawitz converter gas (LDG) draws greatly attention for which the utilization has a close relationship with the energy saving for the enterprise. And accurate prediction for the gas tank levels can be deemed as scientific guidance for energy scheduling and optimization operations. Considering a LDG system of a steel plant in China, a granular computing-based multi-output long-term prediction intervals (PIs) method is proposed to estimate the future trend of gas tank levels. A hybrid collaborative fuzzy clustering model is constructed which utilizing horizontal and vertical structure for fully describes both the relationship between the tank and the related influence factors, and also the mutual impact of the multiple tank levels. Furthermore, by means of the longitudinal expansion on the clustering centers, the prediction horizon is extended from the pointwise mode to the interval values. Then after the optimization along with fuzzy inference and defuzzification, a multi-output long-term PIs construction on the LDG gas tank levels is finally addressed. The results of the experiments and on-site system application indicate the effectiveness and practicability of the proposed method.
With the continuous development of intelligent vehicle technology, automated driving has become a major trend in the future automotive industry. In this study, an electromagnetic tracing vehicle electronic control system based on AURIX TC264D is designed to achieve the speed variability of the trolley in multi-mode scenarios, such as straight paths, roundabout, curve and obstacle avoidance, etc. The multi-mode possibility of the electromagnetic tracing vehicle is verified by constructing the actual scenarios as well as the self-developed electronic control system. The results show that the electronic control system possesses high sensitivity, stability and adaptability to different race tracks, providing reliable support for further research and application in the field of intelligent vehicles.
Ensuring the safety of industrial workers can effectively improve production efficiency. However, there is a lack of methods for recognizing multi-objective behaviors of industrial workers in complex environments, and the accuracy is relatively low. To address these issues, this paper proposes an industrial unsafe behavior recognition network based on target recognition tracking network + VideoMAE to identify unsafe behaviors of industrial workers.First, an industrial field worker action behavior dataset is generated based on collected industrial surveillance videos. Then, a target recognition tracking network is used to accurately locate targets in the multi-objective action behavior dataset. Next, a suitable action behavior recognition network is chosen, combined with object detection and tracking networks, for action behavior recognition. Experimental results demonstrate that the proposed network performs well in recognizing the position and behavior of targets in multi-objective videos under complex conditions such as low light and fog. Compared to other multi-objective recognition networks, it achieves higher accuracy. Therefore, it effectively solves the problem of identifying multi-objective unsafe behaviors of workers in the industrial field.
In this paper, the trajectory tracking problem of the quadrotor UAV with fractional order characteristics is studied. First, an accurate fractional order mathematical model is innovatively established, addressing the deficiency of traditional integer order models in modeling precision. Secondly, the prediction-correction algorithm is applied to solve these complex fractional order differential equations. In design of the control system, the adaptive fuzzy fractional order PID (AFFOPID) control strategy is applied to the position subsystem and attitude subsystem of the quadrotor UAV, dealing effectively with the fractional order characteristics and also enhancing control flexibility. Since the AFFOPID controller does not require the integral term of the error as input, it can be more easily applied and tuned in practical situations. Finally, comparative simulation experiments are conducted to confirm the effectiveness of the proposed control strategy and the superiority of fractional order controllers.
One of the key concerns in today's steel business is how to effectively save energy and reduce emissions, since the concepts of green environmental preservation, energy conservation, and emission reduction are deeply ingrained in people's hearts. Fuel ratio is a key parameter in blast furnace steelmaking. Identifying the factors that influence the fuel ratio and being able to model and predict it can greatly help in guiding the stable operation of the blast furnace, reducing fuel ratio, saving energy, and cutting emissions. In this paper, the fuel ratio mechanism of blast furnace is analyzed, and the parameters affecting the fuel ratio are initially selected. Subsequently, the parameters with the highest correlation to the fuel ratio are identified using the maximum correlation coefficient method. A CNN-LSTM-Attention-based fuel ratio prediction model was developed, and its accuracy and efficacy were demonstrated through comparison with three other models.
