Table of Content

30 September 2023, Volume 44 Issue 9

Previous Issue    Next Issue

Electronic edition of this issue

Electronic edition of this issue

2023, 44(9):  0. 

Asbtract ( )   PDF (137068KB) ( )  

Related Articles | Metrics

Contents

Contents

2023, 44(9):  0. 

Asbtract ( )   PDF (1003KB) ( )  

Related Articles | Metrics

Overview of Key Technology and Its Development of Drone Swarm

LI Jun, CHEN Shichao

2023, 44(9):  2533-2545.  doi:10.12382/bgxb.2023.0514

Asbtract ( )   HTML ( )   PDF (2925KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Modern warfare modes have given birth to the drone swarms as a new type of operational pattern. The concept of forming the advanced group intelligent behavior obtained by using drones with simple functions, which applies the natural bee swarm organization algorithm,is systematically analyzed. The classification and characteristics of drone swarm are introduced. The key technologies and development status of collaborative networking, sensing, decision-making and control are summarized. Focusing on the practical requirements under complex countermeasure environments, the key points and difficulties in the development of unmanned drone swarm technology which integrates electro-optic detection, communication networking, collaborative control and task planning into one are presented. The development approaches and measures of the drone swarm technologies are also put forward, hoping to achieve the goal of leading the development direction of unmanned drone swarm technology and promoting the corresponding technologies to practical applications.

A Review of Edge Computing Technology for Unmanned Swarms

XUE Jianqiang, SHI Yanjun, LI Bo

2023, 44(9):  2546-2555.  doi:10.12382/bgxb.2022.1209

Asbtract ( )   HTML ( )   PDF (3862KB) ( )  

Figures and Tables | References | Related Articles | Metrics

In unmanned swarm operations in the future intelligent warfare, the cloud-edge-end supply of computing power will be an important mode, and its edge computing technology, as the key enabling technology, can solve the problems of poor real-time performance, limited bandwidth, data security and so on. In this paper, we first expound on the concept and technical connotation of edge computing against the background of unmanned swarms. Secondly, we propose a cloud-edge-end distributed system framework for the tactical edge to realize information interconnection, local and global situational awareness on the battlefield, swarm intelligent decision making and cooperative control of unmanned swarm operations. Then, the key technologies involved in the framework are reviewed, including edge computing framework, edge-cloud collaboration, computing offloading, edge command and control, etc. Finally, the outlooks and summary of the unmanned swarm operations with the edge computing technology are presented, providing some reference for edge tactics in the future intelligent warfare.

A Review of Key Technologies for Cross-domain and Trans-medium of Mobile Robotics

SU Bo, JIANG Lei, LIU Yufei, XING Boyang, LI Yongyao, TAN Senqi, WANG Zhirui

2023, 44(9):  2556-2567.  doi:10.12382/bgxb.2023.0414

Asbtract ( )   HTML ( )   PDF (7591KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Human exploration of terrestrial, aerial, spatial, and aquatic environments has never ceased, and this process involves complex scientific problems related to multi-domain and multi-medium environments.As typical examples of machines capable of crossing different environments, mobile robots are expected to explore broader land, higher sky, farther space, and deeper seabed in the future but this presents a significant challenge for mobile robots to maintain high performance in multi-domain and multi-medium environments.This paper first introduces the scientific fusion of mobile robots in energy, space-time, and information domains, highlighting their increasingly important role in exploring new domains and media.Then, it summarizes the basic theoretical research, applications, and development status of mobile robots in areas such as supporting and traction technology, speed and stability technology, and interaction and collaborative technology.Finally, suggestions are proposed to address the key challenges related to the maneuverability and operational intelligence of multi-domain and multi-medium mobile robots.The aim is to enable these machines to perform tasks that are beyond human reach, and to facilitate the widespread application of robot equipment systems in multi-domain and multi-medium environments.

Key Technologies and Application Prospects of Off-road Legged Robot Swarm System

XU Wei, SU Bo, JIANG Lei, YAN Tong, XU Peng, WANG Zhirui, QIU Tianqi

2023, 44(9):  2568-2579.  doi:10.12382/bgxb.2023.0445

Asbtract ( )   HTML ( )   PDF (5900KB) ( )  

Figures and Tables | References | Related Articles | Metrics

As one of the typical ground mobile robots, the legged robot has continuously improved off-road capability and intelligence level. The in-depth integration of leg-based off-road technology and swarm technology is one of the important trends in the development of unmanned equipment in the future. This paper first analyzes the research status of swarm bionic mechanism, off-road single-legged robot, swarm sensing and cooperation, and integrated demonstration and application, puts forward the technical challenges and scientific problems faced by the off-road legged swarm system, and presents in-depth analysis of the core key technologies to be solved from four aspects: high mobility off-road capability of off-road legged robots, swarm self-organization mechanism and collaborative decision-making and control, swarm collaborative situational awareness, and communication ad hoc networks and application validation. Finally, ideas and suggestions for the development prospects of the off-road legged robots swarm in the military and civilian fields are proposed.

Distributed Adaptive Optimal Cooperative Interception Method for Missile Swarm with Synchronization Error Constraints

LIU Dawei, SUN Jingliang, LONG Teng, HE Jing, WANG Xiaoyue

2023, 44(9):  2580-2590.  doi:10.12382/bgxb.2022.1160

Asbtract ( )   HTML ( )   PDF (1960KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To deal with the problem of poor synchronization for partial unknown nonlinear cooperative guidance system subject to synchronization error constraints caused by unknown maneuvering targets, a feedforward compensation+feedback optimization cooperative guidance framework is constructed. Then, a distributed adaptive optimal cooperative interception guidance method is proposed. In the feedforward compensation part, a nonlinear mapping mechanism is provided by designing a novel barrier Lyapunov function (BLF) for compensating the system effects of synchronization error constraints. To estimate the unknown nonlinear term caused by the maneuvering target, a neural network (NN) observer is built, in which the weight values are updated adaptively for estimating the target's unknown maneuvering disturbance. In the feedback optimization part, the virtual and actual control inputs are designed recursively by using the ADP technique, in which the control input in every backstepping process is transformed into the issue of solving the nonlinear coupled HJB equation. An adaptive critic network is built to solve the optimal cost function of the HJB equation online, in which a residual error based adaptive updating law of critic weight is derived. The stability of the nonlinear cooperative guidance closed-loop system is analyzed and the convergence of cooperative synchronization error is guaranteed theoretically based on the Lyapunov theory. The simulation results show that the synchronization error can be reduced to 0.01s while ensuring the interception precision with the proposed method.

A Brief Analysis of Man-machine Hybrid Intelligence Enhancement Technology for Combat Allocation of Squad

GUO Yonghong, LIU Rui, LIU Cheng, LI Xuguang

2023, 44(9):  2591-2599.  doi:10.12382/bgxb.2022.1123

Asbtract ( )   HTML ( )   PDF (2374KB) ( )  

Figures and Tables | References | Related Articles | Metrics

In view of the problems of the traditional combat disposition method, such as extensive planning granularity, poor dynamic adjustment ability and complex improvisation adjustment process,focus on the combat squad level,the potential of artificial intelligence technology to solve the above problems is analyzed and the path selection of man-machine cooperation relationship as “Man-in-the-loop” is given. On this basis,in order to realize the dynamic planning of team assignment and task-oriented flexible adaptation, this paper presents an architecture of combat allocation system based on man-machine hybrid intelligence, and designed its function mode,which is parallel reasoning and optimal recommendation based on the military metaverse. At the same time, the process of by which it works and the expected ability are analyzed. The simulation results show that the proposed method is effective in the decision-making process of task matching, function combination and mapping.The relevant research results can provide theoretical support and technical foundation for the research and engineering construction of intelligent combat allocation strategy and commanding decision-making system oriented to squad level.

Enhanced Multi-scale Target Detection Method Based on YOLOv5

HUI Kanghua, YANG Wei, LIU Haohan, ZHANG Zhi, ZHENG Jin, BAI Xiao

2023, 44(9):  2600-2610.  doi:10.12382/bgxb.2022.1147

Asbtract ( )   HTML ( )   PDF (9899KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To address the problem that the initial anchor box is difficult to match the target and its multi-scale detection ability is not strong in complex scenes, an enhanced multi-scale target detection method based on YOLOv5 is proposed. Through the Kmeans++ clustering algorithm, the multi-scale initialization anchors suitable for the current detection scene is obtained, which makes it easier for the network to capture targets with different scales; then, a number of parallel convolution branches with different scales are added to the Bottleneck structure. While retaining the original feature information, the multi-scale feature information is fused to enhance the global perception ability of the model. The EM-YOLOv5s model proposed is tested on VisDrone2019, COCO2017, and PASCAL VOC2012 datasets. The experimental results show that: compared with the YOLOv5s model, the key indicators such as mAP@0.5∶0.95 and mAP@0.5 are improved; on PASCAL VOC2012, mAP @0.5∶0.95 is increased by 5.2%, while the detection time is only increased by 1.9ms, indicating that EM-YOLOv5 model can effectively improve the target detection accuracy in general complex scenes.

Analysis of Soft Intelligent Edge Computing Technologies for Unmanned Systems

ZHANG Kaige, LU Zhigang, NIE Tianchang, LI Zhiwei, GUO Yuqiang

2023, 44(9):  2611-2621.  doi:10.12382/bgxb.2022.1166

Asbtract ( )   HTML ( )   PDF (1516KB) ( )  

Figures and Tables | References | Related Articles | Metrics

The design and deployment of light-weight neural network models are crucial for intelligent weapon systems. This paper explores the application of intelligent edge computing in unmanned systems from the perspective of building a light-weight deep neural network model, specifically focusing on parameter pruning, knowledge distillation, and parameter quantization techniques. Recent advancements in these fields are discussed, and the performance of various intelligent edge computing technologies is evaluated using object recognition as an example. Based on the advantages and disadvantages of each light-weight design method, a new framework for edge computing is proposed. With improvements in parameter quantization accuracy and introduction of knowledge distillation, the proposed framework becomes feasible for implementation. This approach provides valuable insights for the utilization of intelligent edge computing technologies in enhancing the military intelligence of unmanned systems.

Illumination-aware Multispectral Fusion Network for Pedestrian Detection

PENG Peiran, REN Shubo, LI Jianan, ZHOU Hongwei, XU Tingfa

2023, 44(9):  2622-2630.  doi:10.12382/bgxb.2022.1114

Asbtract ( )   HTML ( )   PDF (6314KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Multispectral pedestrian detection has been widely applied in scenarios such as intelligent security and autonomous driving. However, the accuracy and robustness of pedestrian detection still face challenges, especiallyin low-light conditions or in scenarios with occlusions. To address this issue, a novel pedestrian detection network is proposed, which is namedillumination-aware cross-spectral fusion network. Thenetwork leverages cross-attention and illumination-aware mechanisms to fully exploitmulti-spectral specific features, thereby improving the robustness and accuracy of pedestrian detection. To enhance feature representation between the two spectra, a cross-attention module is introduced. Additionally, an illumination-aware sub-network is proposed, which adaptively selects effective spectral feature information based on the illumination intensity variations of visible and infrared spectra, thusimproving the robustness of the detection system. Experiments areconducted on two multi-spectral pedestrian detection datasets, the KAIST dataset and the CVC-14 dataset. The experimental results demonstratethat theproposed method outperforms existing methods in terms of detection accuracy and speed. This achievementis of significant importance for enhancing the robustness and versatility of pedestrian detection models,with broad potential for practical applications.

Application of Cross-modality Person Re-identificaton based on Edge Intelligent Terminal

BIAN Ziyang, XU Tingfa, MA Liang, LI Jianan

2023, 44(9):  2631-2638.  doi:10.12382/bgxb.2022.1113

Asbtract ( )   HTML ( )   PDF (5419KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Cross-modallity person re-identification technology (cm-ReID) aims to identify the same person in visible and infrared images. It has crucial applications in intelligent systems and equipment for human-machine cooperation, Internet of everything, cross-border integration, and intelligence of everything. In this paper, we propose a cross-modality person re-identification method based on data enhancement, which preserves the structural information of the visible image while performing data enhancement in the wavelength domain to bridge the gap between different modalities. On this basis, a set of edge intelligent terminals are designed and implemented based on the RK3588 chip and a cross-modality person re-identification algorithm is deployed. In the hardware design and software development of edge computing deployment, the system is flexible and scalable through modular design and hierarchical configuration, reducing the computing pressure of centralized data processing. The experimental results show that the proposed method has performed well on two benchmark data sets, SYSU-MM01 and RegDB, and can be deployed in practical scenarios.

Semi-supervised Hyperspectral Salient Object Detection Using Superpixel Attention and Siamese Structure

QIN Haolin, XU Tingfa, LI Jianan

2023, 44(9):  2639-2649.  doi:10.12382/bgxb.2022.1162

Asbtract ( )   HTML ( )   PDF (8180KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Hyperspectral salient object detection technology plays a key role in various fields, such as camouflage recognition and anomaly detection, thus having received extensive attention. The neural network model based on deep learning technology has improved issues such as low detection accuracy and weak robustness of traditional algorithms, but the cost of data labeling limits its further development. To this end, a superpixel attention siamese semi-supervised algorithm is proposed, which uses a small amount of fully supervised data and a large amount of weakly supervised data for training, effectively reducing annotation costs. The algorithm consists of a siamese prediction module and an attention assistance module. The siamese prediction module captures the implicit constraints of weak labels and generates a saliency result map, while the attention assistance module optimizes the prediction results with a superpixel-level channel attention mechanism. The newly proposed semi-supervised algorithm achieves a detection accuracy of 87% on hyperspectral datasets, outperforming other popular algorithms and demonstrating excellent saliency detection performance while effectively reducing annotation costs.

Multi-beam Scanning of Liquid Crystal Optical Phased Array Based on Greedy Algorithm

YE Wenyu, WANG Chunyang, YU Jinyang, TUO Mingkan, WANG Zishuo

2023, 44(9):  2650-2660.  doi:10.12382/bgxb.2022.1163

Asbtract ( )   HTML ( )   PDF (7685KB) ( )  

Figures and Tables | References | Related Articles | Metrics

This paper presents a multi-beam scanning method for liquid crystal optical phased arrays (LC-OPAs) based on greedy genetic algorithm to solve issues of long scanning periods and low capture efficiency associated with single-beam scanning. Multiple LC-OPAs are utilized to generate independent beams with equal scanning areas. To efficiently scan and track multiple dynamic targets, a greedy genetic algorithm-based multi-beam scanning scheme is proposed. The scheme assigns targets to each beam based on target size, position, and velocity information, optimizing the scanning order to maximize target capture probability. A liquid crystal optical phased array multi-beam scanning platform is built, and the proposed scanning method is simulated and experimentally verified under various target numbers, target movement speeds, and spot radii. The results demonstrate that the proposed method can not only effectively reduces the scanning cycle but also improves the capture efficiency of dynamic targets.

Mechanisms of Group Intelligence Emergence in UAV Swarms

GONG Yuanqiang, ZHANG Yepeng, MA Wanpeng, XUE Xiao

2023, 44(9):  2661-2671.  doi:10.12382/bgxb.2022.1181

Asbtract ( )   HTML ( )   PDF (2369KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Understanding the emergence of autonomous cooperative behavior in UAV swarms is a challenging problem. In this study, an analysis method for studying the autonomous cooperative behavior emergence in multi-agent systems is proposed. The method analyzes three levels of the system: the micro-individual layer, the meso-structure layer, and the macro-network layer. It also quantifies the dynamic evolution process of the system from the bottom up, revealing the internal logic of the system from micro to macro and identifying problems in system evolution. The computational experimental model of UAV swarms is also introduced, and the information network of UAV swarm association is designed according to the key features of swarm operations. The game mechanism for public goods is introduced, and the cooperative evolution model of swarms is constructed. The evolution dynamics of swarms on the association network is revealed. Through numerical simulation, the emergence of swarm cooperative behaviors is quantitatively analyzed from different levels of the UAV swarm system, so as to understand the emergence mechanisms of autonomous and cooperative swarm behaviors and provide decision support for the optimization of UAV swarm cooperative mechanisms.

Task Allocation Method of UAV Clusters Based on Sequence Generative Adversarial Network

YAN Yuwen, BI Wenhao, ZHANG An, ZHANG Baichuan

2023, 44(9):  2672-2684.  doi:10.12382/bgxb.2022.0931

Asbtract ( )   HTML ( )   PDF (5305KB) ( )  

Figures and Tables | References | Related Articles | Metrics

In response to the problem that the existing UAV cluster task allocation algorithm decreases the solution efficiency and increases the solution time significantly when performing larger scale task allocations, a task allocation method based on sequence generative adversarial network is proposed. A sequence generation model containing a battlefield information feature extraction network and a sequence generation network are constructed to solve the problem of generating a sequence from battlefield information to task allocation. A discriminative model based on a multicore-multilayer convolutional network is constructed, and a gain-evaluation dual-guided policy gradient update is proposed for model training, which solves the problem of discrete task allocation sequences and ensures the quality of task allocation sequences. Simulation results show that the proposed method can efficiently generate task allocation sequences corresponding to battlefield information while guaranteeing the quality.

Joint Trajectory Planning for Multiple UAVs Target Tracking and Obstacle Avoidance in a Complicated Environment

ZHAO Junmin, HE Haozhe, WANG Shaoqi, NIE Cong, JIAO Yingjie

2023, 44(9):  2685-2696.  doi:10.12382/bgxb.2022.0525

Asbtract ( )   HTML ( )   PDF (5959KB) ( )  

Figures and Tables | References | Related Articles | Metrics

In scenarios where multiple UAVs need to collaborate in ground target tracking tasks within obstacle-dense environments, the obstacle avoidance ability may be insufficient. To address this challenge, we propose a joint trajectory planning algorithm for multiple UAVs, enabling them to simultaneously track targets and avoid obstacles using the null-space method. First, Lyapunov guidance vector field is used to obtain the target tracking velocity command for UAVs when they perform standoff tracking to the ground target coordinately. Obstacle model and artificial potential field function for obstacle avoidance are established, and the obstacle avoidance velocity command for UAVs is obtained by using artificial potential field method. Second, based on the null-space method, the obstacle avoidance task is set as a high-priority task, and the integrated UAV velocity command is obtained through the joint trajectory planning method, which projects the target tracking velocity command into the null-space of the obstacle avoidance task and then adds it to the obstacle avoidance velocity command. Through simulation analysis, the effectiveness of the proposed method is verified. Simulated results show that the proposed joint trajectory planning method can plan effective trajectories for multiple UAVs in real time in complex environments with dense obstacles, and ensure that UAVs avoid dense obstacles and maintain target tracking with good coordination between UAVs.

Distributed Task Allocation Algorithm for Multiple Unmanned Aerial Vehicle Based on Information Retransmission and Package Loss Compensation

CAO Yan, LONG Teng, SUN Jingliang, ZHOU Yuze

2023, 44(9):  2697-2708.  doi:10.12382/bgxb.2022.1180

Asbtract ( )   HTML ( )   PDF (3929KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To address issues such as rapid convergence and communication burden caused by communication package loss, a distributed task allocation algorithm based on information retransmission and package loss compensation is proposed. An inter-UAV information retransmission mechanism is designed. Each UAV adjusts the number of task information retransmissions according to the communication quality, effectively reducing the probability of package loss. The convergence iteration rule of the allocation algorithm with the retransmission mechanism is raised, and the acceleration effect of the retransmission mechanism on the algorithm's convergence is proved. The loss estimation distributed task allocation (LE-DTA) algorithm is proposed to further reduce communication redundancy. The estimation information is used to compensate for lost packages to participate in the distributed task allocation process, which lowers the communication requirements between UAVs. The convergence of the LE-DTA algorithm is then proved. The simulation comparison results show that the improved Consensus-Based Bundle Algorithm (CBBA) with the information retransmission mechanism can effectively accelerate the convergence process, but the disadvantage is the high communication load between UAVs, while the LE-DTA has obvious performance advantages in high package loss rates and low network connectivity scenarios.

Design and Implementation of Hardware-in-the-loop Test Architecture for UAV Information Attack Based on QualNet

LI Siqi, GONG Peng, SHAN Dan, LI Jianfeng, LIU Yu, GAO Xiang

2023, 44(9):  2709-2721.  doi:10.12382/bgxb.2022.1216

Asbtract ( )   HTML ( )   PDF (18700KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To meet the requirements of data link interference attack effect simulation in UAV technology research, a hardware-in-the-loop test architecture for UAV information attack is proposed based on QualNet simulation software. The hardware-in-the-loop test architecture simulates the dynamic control, state information display, and jamming attack loading in UAVs. It integrates data transmission interface and various modules including communication network simulation, scenario planning, jamming attack simulation loading, three dimensional situation display, and dynamic control modules. Test results demonstrate that the architecture can effectively simulate the interference attack effects on UAVs. The designed command data interaction interface ensures accurate and real-time communication in the hardware-in-the-loop test for UAV information attacks, laying a foundation for physical equipment access testing.

An Anti-decoy Jamming Method of Ship-radiated Noise Signals Based on Unsupervised Time-frequency Information Fusion

DUAN Yichen, SHEN Xiaohong, WANG Haiyan, YAN Yongsheng

2023, 44(9):  2722-2731.  doi:10.12382/bgxb.2023.0395

Asbtract ( )   HTML ( )   PDF (4755KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Ship radiated noise signals gives one of the important ways to perceive the target ship. Decoy signals confuse the perception system by imitating the target ship radiated noise to cover the target ship and complete the strategic goal. If anti-decoy jamming can be realized through the recognition of ship radiated noise signals and decoy signals, the efficiency of enemy ship identification in wartime can be greatly improved, so as to improve the efficiency and success rate of tactical operations. The anti-decoy jamming problem is transformed into a one-class classification problem, and the deep learning method is attempted to propose a solution. Against this background, this paper proposes an unsupervised time-frequency information fusion method for anti-decoy jamming of ship radiated noise signals. The generative adversarial network structure for the time domain and time-frequency domain data of the ship radiated noise signal is constructed. The adversarial training strategy is used to capture the time domain and time-frequency domain information of the ship radiated noise signal, which improves the representation learning ability of the model. Finally, the end-to-end one-class classification task of ship radiated noise can be realized. The experimental data are collected from the decoy signals generated by the simulation of the external field experiment. The AUC in the unsupervised condition is 0.84, which is 0.17 higher than the baseline model for one-class classification. The experimental results show that the method can achieve anti-decoy jamming for the ship radiated noise signal.

Optimization Algorithm of Autonomous Target Recognition for Unmanned Vehicles Based on YOLOv5

ZHAO Xiaodong, ZHANG Xunying

2023, 44(9):  2732-2744.  doi:10.12382/bgxb.2022.1161

Asbtract ( )   HTML ( )   PDF (5815KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Autonomous target recognition is a key technology for enabling intelligent operation of unmanned vehicles in complex ground battlefield environments. However, deploying and applyingdeep learning-based autonomous target recognition algorithmson resource-constrained embedded computing platformsfor realizing intelligent unmanned vehicles remains a challenging task. Based on the YOLOv5 deep neural network structure used for autonomous recognition of ground targets, this paper proposes a multi regularization adaptive network pruning algorithm based on animproved attention module and BatchNorm layer. The proposed algorithm achieves optimal pruning of the network structure during the collaborative process of pruning and training. A combined posttraining INT8 quantization algorithm is designed, employing unsaturated mapping for weights and saturated mapping for activation values. The compressed and optimized YOLOv5x network is then deployed and verified on the embedded computing platform based on the Zynq UltraScale+ MPSoC architecture.The verification results show that when YOLOv5x network prunes 40% of the channel and quantizes with INT8 strategy, the recognition accuracy for infrared dataset is only reduced by 0.374%. The recognition accuracy for visible light dataset is increased by 0.065%, and the target recognition frame rate can reach 25 frames per second. The optimized network can meet the accuracy and real-time requirements of autonomous target recognition in the complex battlefield environment of unmanned vehicles. The proposed network optimization algorithm can be extended to other combat platforms, such as unmanned aerial vehiclesand precision-guided weapons.

Firing Data Calculation Method for Laser Terminal Guidance Projectile Based on Multi-model Network

LIU Chang, LEI Hongbo, LIN Shiyao, FAN Shipeng, WANG Jiang

2023, 44(9):  2745-2755.  doi:10.12382/bgxb.2022.1133

Asbtract ( )   HTML ( )   PDF (4199KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To deal with the large truncation errors and defects of being unable to give quick and accurate solutions in calculating the firing data of laser terminal guidance projectiles by the firing table, a method based on multi-model network is proposed. Firstly, a six-degree-of-freedom trajectory model of the guided projectile is established, and the influence of the rear sight on the range in the initial conditions is analyzed. Secondly, considering the influence of air temperature, air pressure, wind interference and other factors on the solution, the least square method is used to fit the air temperature and pressure, and the layer weight method is adopted to obtain the accurate ballistic wind as part of the input of the network model. Thirdly, samples are generated in batches through simulations, which are applied to train each model in the deep neural network. Finally, the proposed method is verified by simulations and flight tests. The simulation results show that the proposed method has a higher precision and an error rate of less than 0.2% compared with the traditional firing table.

Adaptive Attitude Control of Wheel-legged Mobile Platform Based on Feedforward Compensation

LIU Hui, LIU Baoshuai, LIAO Dengting, HAN Lijin, CUI Shan

2023, 44(9):  2756-2767.  doi:10.12382/bgxb.2022.1103

Asbtract ( )   HTML ( )   PDF (5866KB) ( )  

Figures and Tables | References | Related Articles | Metrics

When passing through complex obstacles, a wheel-legged mobile platform bears a relatively large load on the wheel end, and the external force acting on the tire from the ground undergoes sudden changes. This significantly reduces the precision of the platform's attitude control and can lead to tire instability and loss of contact with the ground. To improve the terrain adaptation and stability, an adaptive attitude control strategy for the platform based on feedforward compensation is proposed. Considering the vertical support force and longitudinal driving force at the wheel-ground contact point, the inverse kinematic model and dynamic model of the platform are constructed. And the real-time estimation of the wheel-ground contact state is achieved, and the leg height observer and wheel-ground contact state are combined to perform feedforward compensation to adjust the leg's vertical height, balancing the platform's wheel motion stability and adaptive attitude control accuracy. Furthermore, considering the momentum and angular momentum of the platform, the virtual driving force at the wheel end is optimized by the quadratic programming algorithm to solve the feedforward compensation torque and thus enable the precise control of platform motion. The simulation results show that the proposed method can improve the adaptive attitude control accuracy and tire driving stability of the wheeled-legged mobile platform, laying the foundation for its performing reconnaissance and other tasks in complex working conditions.

A Self-adaptive Dual Radius Filtering Algorithm Based on LiDAR Point Cloud

LIU Bin, LI Xuemei

2023, 44(9):  2768-2777.  doi:10.12382/bgxb.2022.1093

Asbtract ( )   HTML ( )   PDF (6865KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Point cloud denoising is a key step for intelligent driving vehicles to perceive the surrounding environment information. To solve the problem of low operation speed of high-precision denoising in the LiDAR point cloud denoising method, a self-adaptive dual radius filtering method is proposed for complex scenes and multi-scale noise. The 3D point cloud is first simplified by voxel filtering under the constraint of the minimum number of points, and the outliers are preliminarily filtered. Then KD-tree is used to build an index to calculate the average density of point clouds.The adaptive large- and small-radius models are constructed according to the point cloud density to filter drift noise voxels. To verify the effectiveness of the algorithm, in the simple and complex scenes with multiple noise types, the noise removal accuracy and operation speed are compared with other algorithms. In the case of slightly reduced noise removal accuracy, the operation time is less than 0.6 seconds in simple scenes and less than 2 seconds in complex scenes. The new algorithm has high noise removal accuracy and operation speed, as well as a wide range of applications.

Three-dimensional Adaptive Sliding Mode Cooperative Guidance Law with Impact Time and Angle Constraints

WANG Yuchen, WANG Wei, LIN Shiyao, YANG Jing, WANG Shaolong, YIN Zhao

2023, 44(9):  2778-2790.  doi:10.12382/bgxb.2022.1086

Asbtract ( )   HTML ( )   PDF (3884KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Aiming at the problem of three-dimensional multiple missiles intercepting a maneuvering target simultaneously, a three-dimensional distributed cooperative guidance law considering time and space constraints. The proposed guidance law is based on fixed-time stable theory and adaptive sliding mode theory. Along the light-of-sight (LOS) direction, an improved robust adaptive cooperative guidance law is presented based on fixed-time consensus, which can drive a group of missiles to attack maneuvering target at a desired impact time without inherent chattering of sliding mode control. Along the direction vertical to the line-of-sight, a robust adaptive guidance law is proposed based on nonsingular fixed-time terminal sliding mode technique, which can drive missiles attack target with desired impact angle and avoid collision between missiles. Through the Lyapunov stability theory, the fixed-time stability of closed-loop system is demonstrated. The simulation results show that the proposed approach can be implemented to intercept maneuvering targets with desired impact time and impact angle.

Selection and Parameter Optimization of Hybrid Energy Storage System for Intelligent Unmanned Vehicles

HE Qiang, LIU Hougang, ZOU Bo, LÜ Bu, CHEN Xulin, DUAN Yu

2023, 44(9):  2791-2801.  doi:10.12382/bgxb.2022.1074

Asbtract ( )   HTML ( )   PDF (3551KB) ( )  

Figures and Tables | References | Related Articles | Metrics

The rapid development of intelligent unmanned vehicles and their power demands for multi-task loads, along with high-power charging and discharging requirements under complex driving conditions, call for new energy storage system schemes that address the shortcomings of existing single-type energy storage systems. Based on the performance constraints of intelligent unmanned vehicle, such as power performance and silent mileage, a parameter matching optimization method for hybrid energy storage system (HESS) with a focus on lightweight design is proposed to achieve a balance between the input and output capacity of the energy storage system and the overall system quality. By analyzing the basic requirements of unmanned vehicles for energy storage system and the characteristics of different topological structures, the optimal configuration is selected. Based on the requirements of vehicle performance indicators, the energy storage system matching calculation is carried out, and the system parameters are optimized to give full play to the advantages of the hybrid system. The results show that HESS can effectively reduce the high current impact of the power battery and prolong its service life. Meanwhile, the optimization method can effectively reduce the weight of the energy system and improve the comprehensive performance of the vehicle.

Robust Tracking of Quadrotor UAVs Based on Integral Reinforcement Learning

YANG Jiaxiu, LI Xinkai, ZHANG Hongli, WANG Hao

2023, 44(9):  2802-2813.  doi:10.12382/bgxb.2022.1051

Asbtract ( )   HTML ( )   PDF (4758KB) ( )  

Figures and Tables | References | Related Articles | Metrics

A novel robust trajectory tracking control method based on integral reinforcement learning is proposed for the quadrotor UAV position trajectory tracking control with uncertain system model dynamics and external disturbances. Firstly, an augmented system of the original system and reference trajectory of the quadrotor UAV is established to transform the robust trajectory tracking problem of the quadrotor UAV into a sedimentation problem. By using the value function with discount factor, the robust calming problem of the UAV augmented system is transformed into an optimal control problem, taking into account the tracking errors and the overall performance of the quadrotor UAV. Then, based on the integral reinforcement learning method, a single network actor-critic structure is developed to estimate the optimal value function and online solution for the quadrotor UAV controller. Finally, the stability of the quadrotor UAV system tracking errors and the convergence of the single network structure weights are rigorously demonstrated mathematically, and the simulation results verify the superiority and robustness of the proposed control scheme.

Control Barrier Function-based Control for Aircraft Avoidance and Guidance with Dynamic Obstacles

DU Hongbao, WANG Zhengjie, TANG Lixi, ZHANG Xiaoning

2023, 44(9):  2814-2823.  doi:10.12382/bgxb.2022.1002

Asbtract ( )   HTML ( )   PDF (3815KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To improve the engineering applicability of aircraft when pursuing a target in the presence of dynamic obstacles, the relative motion relationship between the aircraft and obstacles, as well as between the aircraft and the target are analyzed. Leveraging the control barrier function (CBF), the velocity obstacle approach is used to constrain the velocity of the aircraft, and a quadratic programming (QP) problem is constructed with the existing proportional guidance law. Then, a closed-form obstacle avoidance guidance law is proposed by solving the QP problem. Numerical simulation and analysis of the trajectory properties and overload show that the proposed obstacle avoidance and guidance law can avoid a moving obstacle in real time and pursue the target with favorable global properties. Compared to existing methods, the overload curve is smoother, and the saturation overload ratio is smaller.

Distributed Target Assignment Method for UAV Swarms Using Identity Hungarian Algorithm

LIU Xingyu, GUO Ronghua, REN Chengcai, YAN Chao, CHANG Yuan, ZHOU Han, XIANG Xiaojia

2023, 44(9):  2824-2835.  doi:10.12382/bgxb.2022.0994

Asbtract ( )   HTML ( )   PDF (3505KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Distributed strike capabilities against multiple enemy targets are crucial for Unmanned Aerial Vehicle (UAV) swarms in combat scenarios. One key challenge is how individual UAVs choose their targets for effective strikes. Most existing target allocation algorithms are designed for centralized target allocation problems with global information, making them unsuitable for battlefield environments with local perception and interaction. To address this, we propose the Identity Hungarian Algorithm, which incorporates drone and target identities into the traditional Hungarian algorithm. This approach considers factors such as proximity, target value, target distance, target azimuth, and UAV speed to achieve distributed target allocation for UAV swarms. Case study results demonstrate that the proposed identity Hungarian Algorithm mitigates target omission and redundancy attacks, enhances the overall combat effectiveness of the UAV swarm, and lays the foundation for effective combat strategies in sequence.

Design and Experimental Study of a Novel Semi-physical Simulation Platform for Visual Navigation of Quadrotor UAVs

HUANG Feng, WANG Weixiong, LIN Zhonglin, WU Xianyu, ZHUANG Jiaquan

2023, 44(9):  2836-2848.  doi:10.12382/bgxb.2022.0760

Asbtract ( )   HTML ( )   PDF (8439KB) ( )  

Figures and Tables | References | Related Articles | Metrics

Visual navigation simulation tests are able to verify the robustness and accuracy of navigation algorithms for Unmanned Aerial Vehicles (UAVs) and expediting the iterative optimization of the algorithms. Traditional computer model-based hardware-in-the-loop simulations fall short in realizing real visual navigation flight tests, so it is necessary to design a high-precision semi-physical simulation platform for UAV visual navigation tests. According to the characteristics of the UAV and the simulation model, a new six-degree-of-freedom mechanical structure composed of a three-axis turntable and a three-axis truss is proposed. This mechanical structure can simulate the flight attitude of a quadrotor UAV within a three-dimensional space of 4.0m×2.0m×1.4m. According to the designed mechanical structure and its dynamic characteristics, the control system based on EtherCAT communication is developed. The system can realize real-time flight attitude simulation in the real physical environment and synchronous flight attitude simulation in the virtual animation space. The actual measurement results show that the repeated positioning accuracy of the three-axis turntable can reach 0.006°, the repeated positioning accuracy of the three-axis truss can reach 0.033mm, and the dynamic error accuracy can reach 0.04° and 0.4mm. The effectiveness of the simulation platform is also verified by indoor and outdoor comparison tests. The results show that the simulation platform can meet the needs of high-precision UAV visual navigation simulation.

UAV Autonomous Air Combat Decision-making Based on AM-SAC

LI Zenglin, LI Bo, BAI Shuangxia, MENG Bobo

2023, 44(9):  2849-2858.  doi:10.12382/bgxb.2022.0669

Asbtract ( )   HTML ( )   PDF (3497KB) ( )  

Figures and Tables | References | Related Articles | Metrics

To address the autonomous decision-making problem of unmanned aerial vehicles (UAV) in modern air combats, a maneuvering decision algorithm based on AM-SAC algorithm is proposed by combining the Attention Mechanism (AM) with Soft Actor Critic (SAC) in deep reinforcement learning. Focusing on 1V1 combat scenarios, the UAV three degree of freedom maneuvering model and the UAV close-range air combat model are established, and the missile attack zone model is built based on the relative distance and relative azimuth angle between both sides in a combat. The attention mechanism is introduced into SAC algorithm to construct the weight network, so as to realize the dynamic adjustment of the weight distribution of reward function during the training process. The simulation experiments are also designed. By comparing with SAC algorithm and testing in multiple environments with different initial situations, it is verified that the UAV air combat decision algorithm based on the AM-SAC algorithm has higher convergence speed and maneuvering stability, as well as better performance in air combat across various initial environments.

Pheromone Positive Incentive Grid Method for Multi-unmanned Platform Regional Surveillance Task

CHEN Yaping, WANG Nan, HONG Huajie, LIU Zhaoyang, YAN Xiangda

2023, 44(9):  2859-2870.  doi:10.12382/bgxb.2022.0537

Asbtract ( )   HTML ( )   PDF (5897KB) ( )  

Figures and Tables | References | Related Articles | Metrics

There is a common problem of regional surveillance in operations in densely populated urban areas. In order to evacuate our side from the highest risk area to reduce damage or manpower consumption, the use of unmanned systems to carry out reconnaissance and surveillance tasks is of great military significance and application value. Aiming at collaborative monitoring tasks with complex and ever-changing environments and multiple unmanned platforms with adjacent initial positions, in response to the shortcomings of existing control strategies that are prone to conflicts when traversing the target space and the lack of research on the proximity of initial positions of multiple unmanned platforms, based on semi heuristic control strategies and grid methods, the objective function is improved by introducing pheromones and developing conflict resolution rules, a pheromone positive incentive grid method is constructed. Experimental results showed that the proposed method performed better in conflict resolution than existing control strategies, its overall performance was better, and the global average idle time was better especially when there were many obstacles. The effectiveness and rationality of the proposed method had been verified.