This paper proposes a novel sparse variant of auto-encoders as a building block to pre-train deep neural networks. Compared with sparse auto-encoders through KL-divergence, our method requires fewer hyper-parameters a...
详细信息
This paper proposes a novel sparse variant of auto-encoders as a building block to pre-train deep neural networks. Compared with sparse auto-encoders through KL-divergence, our method requires fewer hyper-parameters and the sparsity level of the hidden units can be learnt automatically. We have compared our method with several other unsupervised leaning algorithms on the benchmark databases. The satisfactory classification accuracy (97.92% on MNIST and 87.29% on NORB) can be achieved by a 2-hidden-layer neural network pre-trained using our algorithm, and the whole training procedure (including pre-training and fine-tuning) takes far less time than the state-of-art results.
In this paper, several filtering methods for a class of discrete-time stochastic linear time-varying multi-agent systems with local coupling uncertainties have been investigated. Every agent can only observe its own m...
详细信息
In this paper, several filtering methods for a class of discrete-time stochastic linear time-varying multi-agent systems with local coupling uncertainties have been investigated. Every agent can only observe its own measurements (outputs) and its neighbor agents' outputs while the states are invisible to any agent because of communication limitations existing in the considered multi-agent system. Because of the information constraints, without knowing the coupling gains of the local interactions, it is not easy for each agent to estimate its states by traditional Kalman filter or other state observers. Noting of the existence of coupling uncertainties in many practical applications, this paper introduces one general framework of decentralized filtering problem of multi-agent systems. For the considered system, based on the key idea of state augmentation and the certainty-equivalence principle borrowed from principles in adaptive control, we introduce several filtering methods to resolve the fundamental problem considered in this paper. By conducting extensive simulations, the consuming time and estimation errors of every method are compared for one typical example, which suggests which method is more precise and faster.
The computation of stereo depth is a very important field of computer vision. Aiming at solving the problem of low accuracy of traditional Census-based stereo matching algorithm, a variable support-weight approach for...
详细信息
ISBN:
(纸本)9781467321969
The computation of stereo depth is a very important field of computer vision. Aiming at solving the problem of low accuracy of traditional Census-based stereo matching algorithm, a variable support-weight approach for visual correspondence search based on modified Census transform is proposed in this paper. On the basis of analyzing defects of the traditional Census transform, a modified Census transform algorithm using average value of minimum evenness sub-area as a reference instead of the center pixel intensity as a reference is raised which enhances robustness of the algorithm. The matching accuracy is improved by weighting the average value and the standard deviation of Hamming distances in a block. The experiment results indicate that the proposed approach works better than traditional ones. Accurate disparities can be obtained even in the depth discontinuities regions.
This paper is concerned with the state estimation of a linear dynamic system and the fusion of asynchronous multirate multisensor *** dynamic system is described at the finest scale with multiple sensors observing a s...
详细信息
ISBN:
(纸本)9781479900305
This paper is concerned with the state estimation of a linear dynamic system and the fusion of asynchronous multirate multisensor *** dynamic system is described at the finest scale with multiple sensors observing a single target independently at different scales with different sampling *** main result is that the multiscale measurements are properly formulated,and an optimal recursive fusion algorithm is presented.A numerical example is given to show the feasibility and efficiency of the algorithm.
This paper is concerned with the optimal state estimation problem under linear dynamic systems when the sampling rates of different sensors are *** simplicity,we consider two sensors where one's sampling rate is t...
详细信息
ISBN:
(纸本)9781479900305
This paper is concerned with the optimal state estimation problem under linear dynamic systems when the sampling rates of different sensors are *** simplicity,we consider two sensors where one's sampling rate is three times as much as the other'*** noises of different sensors are cross-correlated and are also coupled with the system noise of the previous step. By use of the projection theorem and induction hypothesis repeatedly,a distributed fusion estimation algorithm is *** algorithm is proven to be distributed optimal in the sense of Linear Minimum Mean Square Error(LMMSE) and can effectively reduces the oscillation existed in the sequential ***,a numerical example is shown to illustrate the effectiveness of the proposed algorithm.
Based on the quadruped robot, this paper mainly studies the two directions of the content. The first part mainly introduces the mechanical structure design and the construction of the controlsystem of the quadruped r...
详细信息
Based on the quadruped robot, this paper mainly studies the two directions of the content. The first part mainly introduces the mechanical structure design and the construction of the controlsystem of the quadruped robot, completes the prototype design of the quadruped robot based on hydraulic power system. The second part studies the CPG gait generate method of the quadruped robot based on iterative learning control algorithm. From the principle of bionics, firstly, we use the CPG principle to generate gait, and then use the iterative learning control theory to make the control more optimization.
This paper discusses the design of the mechanical structure of a quadruped robot and a three-dimensional simulation model is built under Adams simulating the mechanism's motion to capture the dynamic characteristi...
详细信息
ISBN:
(纸本)9781467329705
This paper discusses the design of the mechanical structure of a quadruped robot and a three-dimensional simulation model is built under Adams simulating the mechanism's motion to capture the dynamic characteristics of the quadruped *** paper analyses the effect of the length of legs on the walking structure from different aspects: its movement speed,capability of climbing obstacles and the workspace of its foot *** addition,to achieve the optimization target of balancing the output force of each hydraulic cylinder,the paper establishes a mathematical model under required constraints working out the lever members that accommodated with the *** all these factors,the most satisfied structural parameters of the autonomous quadruped robot are selected.
Quadruped robot is expected to serve in complex conditions such as mountain road,grassland,etc.,therefore we desire a walking pattern generation that can guarantee both the speed and the stability of the quadruped ***...
详细信息
ISBN:
(纸本)9781467329705
Quadruped robot is expected to serve in complex conditions such as mountain road,grassland,etc.,therefore we desire a walking pattern generation that can guarantee both the speed and the stability of the quadruped *** order to solve this problem,this paper focuses on the stability for the tort pattern and proposes trot pattern generation for quadruped robot on the basis of ZMP stability *** foot trajectory is first designed based on the work space *** the ZMP and stability margin is computed to achieve the optimal trajectory of the midpoint of the hip joint of the *** angles of each joint are finally obtained through the inverse kinematics ***,the effectiveness of the proposed method is demonstrated by the results from the simulation and the experiment on the quadruped robot in BIT.
Dynamic image stabilization precision of an optical image-stable device is a key technical ***,a fast dynamic image stabilization precision test system for an optical image-stable device is developed.A large-aperture ...
详细信息
Dynamic image stabilization precision of an optical image-stable device is a key technical ***,a fast dynamic image stabilization precision test system for an optical image-stable device is developed.A large-aperture collimator with a designed cross divisional board is used to simulate the infinity *** image-stable device is installed on the motion simulator with six degrees of freedom which is used to simulate the moving state of the *** CCD camera installed behind the eyepiece lens of the image-stable device acquires images rapidly and in real *** local energy maxima center of the cross light spot can be acquired accurately through the proposed algorithm using the Hessian *** addition,to deal with the CCD non-uniformity,an adaptive non-uniformity correction algorithm based on bi-dimensional empirical mode decomposition is *** actual test results for the proposed method show that the test error of dynamic image stabilization is less than 0.7,and the time for the frame image acquisition and processing is less than 10 ms,which demonstrates the effectiveness of the test system.
The polymer electrolyte membrane(PEM) fuel cell has been regarded as a potential alternative power source,and a model is necessary for its design,control and power management.A hybrid dynamic model of PEM fuel cell,...
详细信息
The polymer electrolyte membrane(PEM) fuel cell has been regarded as a potential alternative power source,and a model is necessary for its design,control and power management.A hybrid dynamic model of PEM fuel cell,which combines the advantages of mechanism model and black-box model,is proposed in this *** improve the performance,the static neural network and variable neural network are used to build the black-box *** static neural network can significantly improve the static performance of the hybrid model,and the variable neural network makes the hybrid dynamic model predict the real PEM fuel cell behavior with required ***,the hybrid dynamic model is validated with a 500 W PEM fuel *** static and transient experiment results show that the hybrid dynamic model can predict the behavior of the fuel cell stack accurately and therefore can be effectively utilized in practical application.
暂无评论