检索结果-南通市图书馆

Accelerated Learning for Restricted Boltzmann Machine with a Novel Momentum Algorithm

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Chinese Journal of Electronics 2018年第3期27卷 483-487页

作者： GAO Xiaoguang LI Fei wan kaifang Northwestern Polytechnical University

We investigated two commonly used momentum algorithms, Classical momentum(CM) and Nesterov momentum(NM). We found that, when used in Restricted Boltzmann machine(RBM), they have two main problems: The first one is their performances are not obvious and not as good as expected. The second one is they may lose accelerating ability in the later stage of training process. Aiming at these two problems, we proposed the Weight momentum algorithm and evaluated our approach on four datasets. It has been demonstrated that our methods can achieve better performance under both reconstruction error and classification rate criterions.

关键词： Deep learning Restricted Boltzmann machine(RBM) Momentum algorithm Weight momentum(WM)

Training Restricted Boltzmann Machine Using Gradient Fixing Based Algorithm

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Chinese Journal of Electronics 2018年第4期27卷 694-703页

作者： LI Fei GAO Xiaoguang wan kaifang Northwestern Polytechnical University

Most of the algorithms for training restricted Boltzmann machines(RBM) are based on Gibbs sampling. When the sampling algorithm is used to calculate the gradient, the sampling gradient is the approximate value of the true gradient and there is a big error between the sampling gradient and the true gradient, which seriously affects the training effect of the network. Aiming at this problem, this paper analysed the numerical error and orientation error between the approximate gradient and the true gradient. Their influence on the performance of network training is given then. An gradient fixing model was established. It was designed to adjust the numerical value and orientation of the approximate gradient and reduce the error. We also designed gradient fixing based Gibbs sampling training algorithm(GFGS) and gradient fixing based parallel tempering algorithm(GFPT), and the comparison experiment of the novel algorithms and the existing algorithms is given. It has been demonstrated that the new algorithms can effectively tackle the issue of gradient error, and can achieve higher training accuracy at a reasonable expense of computational runtime.

关键词： Deep Learning Restricted Boltzmann machine(RBM) Gradient fixing Gibbs sampling training algorithm(GFGS) Gradient fixing based parallel tempering algorithm(GFPT)

Placement of unmanned aerial vehicles as communication relays in two-tiered multi-agent system:clustering based methods

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Journal of Systems Engineering and Electronics 2020年第2期31卷 231-242页

作者： WU Gaofeng wan kaifang GAO Xiaoguang FU Xiaowei School of Electronics and Information Northwestern Polytechnical UniversityXi’an 710129China

The network performance and the unmanned aerial vehicle(UAV)number are important objectives when UAVs are placed as communication relays to enhance the multi-agent information *** problem is a non-deterministic polynomial hard(NP-hard)multi-objective optimization problem,instead of generating a Pareto solution,this work focuses on considering both objectives at the same level so as to achieve a balanced solution between *** on the property that agents connected to the same UAV are a cluster,two clustering-based algorithms,M-K-means(MKM)and modified fast search and find density of peaks(MFSFDP)methods,are first *** the former algorithm requires too much computational time and the latter one requires too many relays,an algorithm for the balanced network performance and relay number(BPN)is proposed by discretizing the area to avoid missing the optimal relay positions and defining a new local density function to reflect the network performance *** results demonstrate that the proposed algorithms are feasible and *** between these algorithms show that the BPN algorithm uses fewer relay UAVs than the MFSFDP and classic set-covering based algorithm,and its computational time is far less than the MKM algorithm.

关键词： unmanned aerial vehicle(UAV) relay communication clustering relay node placement wireless network

Relevant experience learning:A deep reinforcement learning method for UAV autonomous motion planning in complex unknown environments

在线全文

学校读者我要写书评

暂无评论

Chinese Journal of Aeronautics 2021年第12期34卷 187-204页

作者： Zijian HU Xiaoguang GAO kaifang wan Yiwei ZHAI Qianglong wanG School of Electronic and Information Northwestern Polytechnical UniversityXi'an 710129China

Unmanned Aerial Vehicles(UAVs)play a vital role in military *** a variety of battlefield mission scenarios,UAVs are required to safely fly to designated locations without human ***,finding a suitable method to solve the UAV Autonomous Motion Planning(AMP)problem can improve the success rate of UAV missions to a certain *** recent years,many studies have used Deep Reinforcement Learning(DRL)methods to address the AMP problem and have achieved good *** the perspective of sampling,this paper designs a sampling method with double-screening,combines it with the Deep Deterministic Policy Gradient(DDPG)algorithm,and proposes the Relevant Experience Learning-DDPG(REL-DDPG)*** REL-DDPG algorithm uses a Prioritized Experience Replay(PER)mechanism to break the correlation of continuous experiences in the experience pool,finds the experiences most similar to the current state to learn according to the theory in human education,and expands the influence of the learning process on action selection at the current *** experiments are applied in a complex unknown simulation environment constructed based on the parameters of a real *** training experiments show that REL-DDPG improves the convergence speed and the convergence result compared to the state-of-the-art DDPG algorithm,while the testing experiments show the applicability of the algorithm and investigate the performance under different parameter conditions.

关键词： Autonomous Motion Planning(AMP) Deep Deterministic Policy Gradient(DDPG) Deep Reinforcement Learning(DRL) Sampling method UAV

Using approximate dynamic programming for multi-ESM scheduling to track ground moving targets

在线全文

学校读者我要写书评

暂无评论

Journal of Systems Engineering and Electronics 2018年第1期29卷 74-85页

作者： wan kaifang GAO Xiaoguang LI Bo LI Fei Key Laboratory of Aerospace Information Perception and Photoelectric Control Ministry of EducationSchool of Electronics and Information Northwestern Polytechnical University

This paper researches the adaptive scheduling problem of multiple electronic support measures(multi-ESM) in a ground moving radar targets tracking application. It is a sequential decision-making problem in uncertain environment. For adaptive selection of appropriate ESMs, we generalize an approximate dynamic programming(ADP) framework to the dynamic case. We define the environment model and agent model, respectively. To handle the partially observable challenge, we apply the unsented Kalman filter(UKF) algorithm for belief state estimation. To reduce the computational burden, a simulation-based approach rollout with a redesigned base policy is proposed to approximate the long-term cumulative reward. Meanwhile, Monte Carlo sampling is combined into the rollout to estimate the expectation of the rewards. The experiments indicate that our method outperforms other strategies due to its better performance in larger-scale problems.

关键词： sensor scheduling target tracking approximate dynamic programming non-myopic rollout belief state

A learning-based flexible autonomous motion control method for UAV in dynamic unknown environments

在线全文

学校读者我要写书评

暂无评论

Journal of Systems Engineering and Electronics 2021年第6期32卷 1490-1508页

作者： wan kaifang LI Bo GAO Xiaoguang HU Zijian YANG Zhipeng School of Electronic and Information Northwestern Polytechnical UniversityXi’an 710072China

This paper presents a deep reinforcement learning(DRL)-based motion control method to provide unmanned aerial vehicles(UAVs)with additional flexibility while flying across dynamic unknown environments *** method is applicable in both military and civilian fields such as penetration and *** autonomous motion control problem is addressed through motion planning,action interpretation,trajectory tracking,and vehicle movement within the DRL *** DRL algorithms are presented by combining two difference-amplifying approaches with traditional DRL methods and are used for solving the motion planning *** improved Lyapunov guidance vector field(LGVF)method is used to handle the trajectory-tracking problem and provide guidance control commands for the *** contrast to conventional motion-control approaches,the proposed methods directly map the sensorbased detections and measurements into control signals for the inner loop of the UAV,i.e.,an end-to-end *** training experiment results show that the novel DRL algorithms provide more than a 20%performance improvement over the state-ofthe-art DRL *** testing experiment results demonstrate that the controller based on the novel DRL and LGVF,which is only trained once in a static environment,enables the UAV to fly autonomously in various dynamic unknown ***,the proposed technique provides strong flexibility for the controller.

关键词： autonomous motion control(AMC) deep reinforcement learning(DRL) difference amplify reward shaping

Imaginary filtered hindsight experience replay for UAV tracking dynamic targets in large-scale unknown environments

在线全文

学校读者我要写书评

暂无评论

Chinese Journal of Aeronautics 2023年第5期36卷 377-391页

作者： Zijian HU Xiaoguang GAO kaifang wan Neretin EVGENY Jinliang LI School of Electronics and Information Northwestern Polytechnical UniversityXi’an 710129China School of Robotic and Intelligent Systems Moscow Aviation Institute(National Research University)Moscow 125993Russia Electromagnetic Space Operations and Applications Laboratory The 29th Research Institute of China Electronics Technology Group CorporationChengdu 610036China

As an advanced combat weapon,Unmanned Aerial Vehicles(UAVs)have been widely used in military *** this paper,we formulated the Autonomous Navigation Control(ANC)problem of UAVs as a Markov Decision Process(MDP)and proposed a novel Deep Reinforcement Learning(DRL)method to allow UAVs to perform dynamic target tracking tasks in large-scale unknown *** solve the problem of limited training experience,the proposed Imaginary Filtered Hindsight Experience Replay(IFHER)generates successful episodes by reasonably imagining the target trajectory in the failed episode to augment the *** welldesigned goal,episode,and quality filtering strategies ensure that only high-quality augmented experiences can be stored,while the sampling filtering strategy of IFHER ensures that these stored augmented experiences can be fully learned according to their high *** training in a complex environment constructed based on the parameters of a real UAV,the proposed IFHER algorithm improves the convergence speed by 28.99%and the convergence result by 11.57%compared to the state-of-the-art Twin Delayed Deep Deterministic Policy Gradient(TD3)*** testing experiments carried out in environments with different complexities demonstrate the strong robustness and generalization ability of the IFHER ***,the flight trajectory of the IFHER agent shows the superiority of the learned policy and the practical application value of the algorithm.

关键词： Artificial intelligence Autonomous navigation control Deep reinforcement learning Hindsight experience replay UAV

Mobility control of unmanned aerial vehicle as communication relay in airborne multi-user systems

在线全文

学校读者我要写书评

暂无评论

Chinese Journal of Aeronautics 2019年第6期32卷 1520-1529页

作者： Gaofeng WU Xiaoguang GAO Xiaowei Fu kaifang wan Ruohai DI School of Electronics and Information Northwestern Polytechnical University

In this paper, a model-based adaptive mobility control method for an Unmanned Aerial Vehicle(UAV) acting as a communication relay is presented, which is intended to improve the network performance in airborne multi-user systems. The mobility control problem is addressed by jointly considering unknown Radio Frequency(RF) channel parameters, unknown multi-user mobility, and non-available Angle of Arrival(AoA) information of the received signal. A Kalman filter and a least-square-based estimation algorithm are used to predict the future user positions and estimate the RF channel parameters between the users and the UAV, respectively. Two different relay application cases are considered: end-to-end and multi-user communications. A line search algorithm is proposed for the former, with its stability given and proven, whereas a simplified gradient-based algorithm is proposed for the latter to provide a target relay position at each decision time step, decreasing the two-dimensional search to a one-dimensional search. Simulation results show that the proposed mobility control algorithms can drive the UAV to reach or track the optimal relay position movement, as well as improving network performance. The proposed method reflects the properties of using different metrics as objective network performance functions.

关键词： Channel estimation Gradient methods Motion control Optimization Relay Unmanned aerial vehicle Wireless networks

Intercept Mode Suitable for the Space-Based Kinetic Energy Interceptor

在线全文

学校读者我要写书评

暂无评论

Journal of Shanghai Jiaotong university(Science) 2019年第5期24卷 671-680页

作者： DUAN Junhong LIU Yuefeng wan kaifang Air and Missile Defense College Air Force Engineering UniversityXi'an710051China Science and Technology onElectro-Optic Control Laboratory Luoyang471000HenanChina School of Electronics and Information Northwestern Polytechnical UniversityXi'an710129China

The intercept mode of space-based kinetic energy interceptor(KEI) is an essential problem to be solved in the concept design of KEI and is closely related to the KEI’s detailed designing and practical ***, this problem has not yet been investigated systematically and comprehensively. A suitable intercept mode for the KEI is proposed in this research. The intercept mode consists of three basic elements: the specific operation procedure, general scheme for the intercept trajectory, and mathematical model of the intercept *** example is provided to demonstrate the feasibility and effectiveness of the intercept mode. The intercept mode proposed in this research is capable of achieving a rapid intercept, and is applicable to the universal space interception.

关键词： kinetic energy interceptor(KEI) intercept mode trajectory planning guidance law pregnable area