An on-line learning neural controller for helicopters performing highly nonlinear maneuvers

This paper presents an on-line learning adaptive neural control scheme for helicopters performing highly nonlinear maneuvers. The online learning adaptive neural controller compensates the nonlinearities in the system and uncertainties in the modeling of the dynamics to provide the desired performance. The control strategy uses a neural controller aiding an existing conventional controller. The neural controller is based on a online learning dynamic radial basis function network, which uses a Lyapunov based on-line parameter update rule integrated with a neuron growth and pruning criteria. The online learning dynamic radial basis function network does not require a priori training and also it develops a compact network for implementation. The proposed adaptive law provides necessary global stability and better tracking performance. Simulation studies have been carried-out using a nonlinear (desktop) simulation model similar to that of a BO105 helicopter. The performances of the proposed adaptive controller clearly shows that it is very effective when the helicopter is performing highly nonlinear maneuvers. Finally, the robustness of the controller has been evaluated using the attitude quickness parameters (handling quality index) at different speed and flight conditions. The results indicate that the proposed online learning neural controller adapts faster and provides the necessary tracking performance for the helicopter executing highly nonlinear maneuvers.     

GemstoneFire: adaptive dispersive ray tracing of polyhedrons

Synthesizing realistic images of gemstones requires techniques beyond the scope of normal ray tracing. The fire of such highly refractive objects is what makes gemstones attractive, and also imposes very high computational overhead to perform time consuming dispersive ray tracing. Gemstones are usually cut in polyhdrons as for example, a brillant cut. After a detailed analysis of the nature of dispersive ray tracing of polyhedral objects, we propose here a new method of using three simple rays adaptively to model the ray spreading caused by dispersive refraction. It is shown that the proposed method reduces the computational complexity to an order close to that of normal ray tracing.     

基于神经网络的聚类方法研究

针对常用聚类方法不能有效处理噪声数据的问题,本文结合神经网络具有自适应性的特点,提出基于神经网络的聚类(NN_Cluster)模型,并设计了基于自适应共振理论的神经网络聚类模型(ARTNN_Cluster)和基于自组织特征映射的神经网络聚类模型(SOMNN_Cluster)。标准数据集上的实验结果表明,与传统的K_means聚类方法相比,本文提出的基于神经网络的聚类模型有效地克服了传统方法的噪声问题,得到了较好的聚类效果。     

基于变论域理论的模糊自适应控制方法研究

针对一类非线性、快时变性以及模型不确定性问题,设计了一种基于变论域的模糊自适应控制器。利用模糊模型逼近控制系统的非线性部分,以及采用李亚普诺夫方法证明所设计的非线性控制系统的渐进稳定性并推导出自适应律。最后将设计的控制器应用于倒立摆,仿真结果表明跟踪性能良好。     

航迹融合算法在多传感器融合中的应用

研究寻的制导优化控制问题,针对传统单一传感器导引不能满足性能要求,提出采用多传感器复合制导.航迹融合是多传感器数据融合中一个非常重要的方面.由于公共过程噪声的原因,使在应用状态估计融合系统中,来自不同传感器的航迹估计误差未必有独立性,为了使航迹与航迹关联和融合,提出自适应航迹和协方差加权航迹融合的算法.通过仿真研究说明自适应航迹融合和协方差加权航迹融合的算法对多传感器数据融合技术有很明显的作用,数据融合效果好,为复合寻的制导优化设计提供了依据.     

医学图像中粘连细胞分割方法研究

研究粘连细胞图像的分割问题。针对医学细胞图像中,细胞经常发生粘连,细胞粘连部分由于发生像素灰度混合,造成图像不清晰。传统的应用灰度的分割算法由于不能很好的区分粘连部分的灰度,不能完整的分割细胞的问题。为了解决上述问题,提出了自适应阀值调整的粘连细胞分割方法。通过求得粘连部位的细胞像素的具体特征,运用阀值的自适应变化区分粘连部位像素之间的不同,对像素进行划分,克服了传统方法像素划分不清的弊端。实验表明,方法能够有效的分割医学图像中的粘连细胞,取得了比较好的效果。     

PID控制器设计与仿真研究

研究优化PID控制器性能问题,在过程控制中,系统存在纯滞后性、时变性、非线性的复杂系统,传统模糊控制和PID控制方法解决上述问题效果不佳,导致超调量大和调节时间长。为了解决控制系统中超调量大和调节时间长等难题,在Smith预估器和传统PID控制的基础上,结合模糊PID控制的优点,提出一种模糊PID-Smith的智能控制方法。方法通过对模型的纯滞后进行补偿,来减小纯滞后系统的超调量,从而增强系统的稳定性。通过对火电机组主汽温控制系统进行仿真,结果表明,方法降低了超调量,减少了调节时间,极大地改善了控制性能,提高了控制系统的自适应性和鲁棒性。     

基于神经网络的助力搬运装置自适应控制方法研究

为提高助力搬运装置在建筑工地等复杂路况下的适用性,通过对国内外助力系统及其控制方法发展现状的调查和分析,提出一种智能助力控制系统。首先建立控制模型,在智能控制的研究中发现BP神经网络能够对输入信息进行识别并分类处理,且计算较简单,容易实现,但BP神经网络隐含层节点数难以确定、易陷入局部最小值,因此,采用遗传算法对BP神经网络进行优化(GA-BP),并设计具体实现方法。采集平坦、斜坡两种路况的数据,通过人工赋予理想输出数据,将其一部分作为训练数据,对GA-BP神经网络及BP神经网络进行对比训练,另一部分作为验证数据,检验两种网络的训练结果。使用MATLAB进行数据仿真分析,验证GA-BP神经网络控制方法的可行性。试验结果表明:这种控制方法能够有效地解决助力搬运装置在复杂路况下的自适应控制问题。     

锯床的数字化改造

为了提高老式带锯床的精度和自动化程度，对锯床的角度测量部分进行了数字化改造。经过改造，锯床的精度得到了提高，可精确地测量和显示转台所在的角度；同时，通过对驱动控制部分的简单设计，可使锯床在键盘输入角度信号后，控制转台自动地转到此位置。     

A Uniform Approach for Utilizing Synergy between Inspection Technologies and Computational Methods

Emerging scanning technologies currently capture diverse data that includes geometrical and physical properties (i.e. normals, material and color). Yet, existing inspection computational methods do not satisfy industry demands. This paper proposes a new approach for inherently handling diverse data and utilizing the synergy between inspection technologies and digital processing methods. Two methods were developed: a Hierarchical Space Decomposition Model (HSDM), which sorts the sampled data, and a 3D Geometrical Bilateral Filter (3D GBF) that utilizes the diverse data. These methods provide efficient data reduction while preserving sharp features. The feasibility of this approach is demonstrated on the diverse data.