摘要
针对机器人环境识别问题,研究其工作环境描述与实现过程,提出一种环境拓扑地图建立的新方法。该方法以自组织特征映射图的工作算法为基础,提出GSOM(Growing Self-organizing Map)算法,该算法具有增长特性,通过不断增加新的神经元实现网络规模的增长,从而满足描述环境特征的需要,建立环境拓扑地图;仿真试验表明GSOM算法的正确性,可以在样本数未知情况下,确定描述环境特征的最优SOM神经元数量,以少数SOM图神经元分布描述具有大量特征信息的环境结构,建立更能准确描述环境的拓扑地图。
The description of the robot＇s work environment and the actualizing process was studied, and a new method of making topologic map of the environment was proposed. Based on the SOM（Self-organizing Map） the GSOM（Growing SOM） was proposed that the size of the neural network was growing in order to adapt the description of the environment. The experiment proves the validity of the algorithm. The algorithm can get the superior number of the SOM neuron without knowing the quantity of the sample, describe the complex environment with much characteristic information using few SOM neurons and build up the more exact topologic map.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2008年第1期81-84,98,共5页
Journal of System Simulation
基金
国家自然科学基金课题(60375017)
北京市人才强教计划项目(05002011200506)
高等学校博士学科点专项科研基金(20050005002)
