Search

This paper investigates model reference control for linear time-varying (LTV) systems. It is shown that the problem can be decomposed into two sub-problems: a state feedback stabilization problem of LTV systems and a feed-forward compensation problem. Firstly, the sufficient condition for the existence of the model reference control is deduced. The condition concerns two coefficient matrices such that two matrix equations are met simultaneously, and the state feedback gain matrix stabilizes LTV systems according to the unified spectral theory. Secondly, generally complete parametrization of the model reference tracking controller including a state feedback controller and a feed-forward compensator is proposed based on the solution to a type of generalized Sylvester matrix equations. Furthermore, a general procedure for solving the model reference control problem is also provided. Finally, a numerical example shows the feasibility and effectiveness of the parametric design approach to the model reference control in LTV systems.

Multi-UAVs swarm is the focus of current research. UWB (ultra-wide band) is very feasible to be applied to UAV(unmanned areial vehicle) relative positioning. The ranging accuracy, refresh frequency and positioning accuracy of UWB are very important to the formation flight of UAV. The paper mainly studies the applicability and main influencing factors of UWB distributed relative positioning for UAV cluster in tight formation flight. First of all, we analyzed the current cluster relative positioning method and UWB application. Meanwhile, we designed a distributed relative measurement system. Then, the measurement error of UWB point to point is studied, the distributed positioning algorithm is designed and the error analysis is carried out. Finally, the influence of relative speed and positioning result of UAV is emphatically studied and analyzed through experiments.

According to the application characteristics of large-scale fixed wing UAV swarm in military and civil fields, this paper firstly analyzes the communication requirements and technical challenges of swarm networking, and summarizes the major technical approaches to swarm communication. On this basis, the ad-hoc network suitable for swarm application is studied, and comparative simulations are performed on MAC protocols (i.e., the centralized /distributed TDMA, CSMA/CA) and the routing protocols OLSR and AOD according to the characteristics of cluster tasks. A design scheme for hierarchical cluster ad-hoc network is proposed and verified by simulation.

The order picking and distribution are two necessary activities in logistics, the cost of which accounts for a considerable portion of the logistics cost. Although there are dependencies between these two activities, a lot of work has been carried out on them independently; hence, the optimization may be compromised. This paper proposes an integrated order batching and distribution model. The objectives are to minimize the makespan (including picking time and delivery time) and to minimize the number of delivery trips. To solve the integrated model, an EDA approach is devised, in which an elaborate encoding/decoding method is employed and an incremental learning probability model is designed. Experiments show that the proposed integrated model and its solution approach can lead to a significant reduction in the total cost (weighted sum of the makespan and the number of delivery trips).

Based on a proposed complete parametric approach for eigenstructure assignment in second-order linear system via the proportional-plus-derivative feedback control, insightful parametrizations of the closed loop eigenvalues sensitivities to the perturbed elements in the open loop system matrices are obtained, and an effective algorithm for eigenvalue assignment with minimum sensitivity in second-order systems via the proportional-plus-derivative feedback control is then proposed, The algorithm does not contain ‘go back’ procedures, and allows the closed loop eigenvalues to be conveniently optimized with desired regions, A numerical example demonstrates its effect, simplicity and numerical property

Few-shot learning, which focuses on solving machine learning problems in the setting of scarce data, has become a hot spot in the field of neural network and machine learning recently. Inspired by BlockDrop that performs pruning compression on the network using reinforcement learning, in this paper, we present a Group-Aware Pruning (GAP) method with new unifying pruning strategies and different training/testing ways to fit few-shot learning. The proposed GAP consists of three modules, that is, a pruning module, a strategy consensus module (SCM), and a classification module. The whole support set is first fed into the pruning module to get a pruning strategy for each image. Next, the strategies are fed into SCM to fuse into a group strategy for further pruning. When the classification module makes a prediction for a query image, it no longer needs re-output the pruning strategy for the query image since pruning strategies have been unified. Note that in SCM, strategy unification is proposed to achieve the group-aware strategy, which assures an existing deep network that has been pruned by the group-aware strategy work well for few-shot learning problems. Additionally, the testing will be speeded up due to the fact that only one classification model is required in the proposed GAP. Experiments on two benchmark datasets, namely, Omingnet and miniImageNet, and comparisons with the existing state-of-the-art methods show that the accuracy of the proposed GAP is 4.94% higher than the state-of-the-art methods on the 5-way 5-shot task, which shows the effectiveness of the proposed method.

It is a challenge to track a moving target with deformation, occlusion or motion blur in a video stream. Various machine learning algorithms have been proposed for target tracking such as TLD (Tracking-Learning-Detection) and MTT (Multi-task Tracking). However, these algorithms still have limitations to detect and recover from tracking failures. In this paper, we propose hMTT that introduces human into the tracking loop of MTT. Specifically, the tasks are represented and separated as related particles and outlier particles. We design a human-guided mechanism that keeps the related particles as transferable knowledge and replaces the outlier particles with more reliable ones whenever a tracking drift or a tracking failure happens. Meanwhile, the multi-task learning module keeps updating the shared features stored as templates in a dictionary. In addition, the overall confidence value of the tracking result can be calculated accordingly, and the confidence level can be learned on-line for result classification. Experimental results have demonstrated that hMTT outperforms TLD and MTT in various tracking tasks.

This paper discusses the dynamic task allocation problem for manned-unmanned aerial vehicle teamwork (MUAVT) in complex missions. We assume the team is heterogeneous that the team members (MAVs or UAVs) have different capacities for specific tasks. Based on the values of capacities, we analyze the interdependence relationships in the joint activities and assign the roles of performer or supporter to every team member. Then, we calculate the success probability of the tasks with consideration of penitential threats. The tasks can be reallocated whenever necessary in case of aircraft failures, change of task priority, or emergence of new tasks. Specifically, we have proposed an auction-based algorithm to deal with dynamic task allocation with constrains of time and resources. Our method has been evaluated in simulation where one MAV and four UAVs collaborate to handle multiple targets in a surveillance task.

Damage assessment is urgently needed for analyzing how badly a building has been destructed by weapons or by other disasters. However, it is dangerous for humans to go to the destructed area and assess the situation in real-time. In this paper, we propose a human-machine collaborative approach for the damage assessment task using UAV images. Specifically, we use the coactive design method to carry out the interdependence analysis of human and machine in each subtask. After matching the scenes before and after the damage, appearance differences of the target building are captured by ratio method. Then, we use a pre-trained decision tree to evaluate the degree of functional damage of the building. We demonstrate the effectiveness of the proposed method with images captured by a UAV in real-world environments.

The design problem about interval observers for linear systems with time-varying disturbances is considered in the paper. With the nonsingular matrix transformation, the problem about solving the observable gain matrix complicatedly or the situation of nonexistent such matrix is efficiently solved by transforming it into the solution for a type of generalized Sylvester equations. Moreover, the specific parametric forms for the observable gain matrix and the nonsingular matrix are obtained by solving the generalized Sylvester equation. Further, a novel parameterized approach to design interval observers for linear systems with time-varying disturbances is proposed, and the output interval can be towards a constant with a desired convergence rate. Finally, a numerical example illustrates the efficiency of the proposed method.

How to improve the efficiency of power system stability and the whole system is the main problem facing the photovoltaic power system, based on the real system experiments and the digital simulation platform for the limitations of the study, this article from the photovoltaic power generation system semi physical simulation research of power system simulation software based on the use of advanced real-time simulation software in the laboratory the establishment of photovoltaic power generation, electromagnetic transient model, and on this basis, using ADPSS platform of photovoltaic power generation system semi physical simulation experiment, The reliability and stability of the grid connected system of the whole PV power station are verified by experiments.

It is an important topic in the current image processing and analysis to use the photos or video to determine the shooting position and time. Analyzing the change of the object's sun shadow is an effective way to solve this problem. In this paper, the 3D scene photography is modeled to simulate the virtual reality by the software UnrealEngine 4 and the scene of the video can be reproduced accurately. The values of the solar angle parameters in the virtual environment can be shown on the software interface directly, so the solar elevation angle and solar azimuth angle can be fitted with these data to simplify the unnecessary mathematical derivation and programming. Furthermore, a measurement that reflects the deviation of the theoretical values calculated by the formula and fitted values is set as Difference of Angle. Finally, with the minimum Difference of Angle as the optimization goal, the rough possible range of the optimal solution is determined by the discretization search and the optimum solution is calculated by the least squares method. The result shows that this model is generally-used and operable. This method based on virtual reality technology has high maneuverability and accuracy, with good scalability for the number of unknown parameters.

This paper proposes a complete parametric approach for eigenstructure assignment in second-order system Mq+Dq+kq=cu using the displacement-plus-acceleration feedback controller. The expressions of closed-loop eigenvector matrix and feedback gains are in direct closed forms in terms of the closed-loop eigenvalues and a group parameter vectors, which represents the degree of design freedom. The main advantage of the proposed method is solved directly in matrix second-order framework, without transformation into the first-order form, and the numerical stability is better because it used only right coprime factorization and simple matrix transformation. The proposed algorithm is simple and does not contain ‘go back’ procedures. Finally, computational results show its effect and simplicity.

This paper presents two iterative selecting threshold method for component surface defect image segmentation based on gray level. Firstly, it makes use of the statistical gray histogram and two iterative selecting threshold method to split the defect area of the component surface image. Secondly, in order to constructing each of the connected region of the defective parts, it adopts morphological closing operation on the image to connecting smaller discontinuous area. Finally, by use of extracting features of the defect region and matching features to detect and identify the pitting defects and scratch defects. Experiments show that the proposed method can effectively detect the main scratch and pitting defects of the component surface image and has fast detection speed, good real-time performance.

When used in counter-terrorism and disaster rescue, UAVs may possibly fly in buildings. In an indoor closed environment, GPS signal may be blocked, which will make UAVs difficult to obtain their location. In addition, a complex indoor environment also brings challenges to the safety of flight. Therefore, it is necessary to identify complex environments and avoid obstacles in real-time. This paper introduces the PTAM algorithm to the UAV ground control module. To deal with poor illumination of indoor environment, less feature points and other issues, we improve the algorithm and design a self-adaptive map feature point generating mechanism, reducing the dependency of the algorithm on the number of feature points and illumination conditions. On this basis, this paper proposes an obstacle recognition algorithm and designs a warning mechanism for the ground control module. Finally, we use a small quad-rotor UAV to test this algorithm. The results show that the proposed method can provide warning to the obstacles during flight in unknown indoor environments with self-localization. The experiments demonstrate that this algorithm is effective in real-time, which has great significance to explore autonomous positioning and flight safety for UAVs in GPS denied environments.