摘要
随着市场对定位需求及精度要求的提高,针对室内复杂环境而导致低精度定位的问题,提出使用BP训练方法取代传统方案进行测距模型的建立,以消除对环境的过度经验依赖,提高算法针对不同环境的普适性;并在此基础上,研究了实测距离与模型预测距离关系,建立锚节点概率环形区间,提出概率区间交叠定位算法,并以多边界质心算法为辅,得到一种新的混合定位方法。该方法在CC2420/Tiny OS平台验证,通过实测试验证明其在定位精度上较传统加权质心算法/质心算法提高了17.2%和37%。
With the growth in demand of localization and the low localization precision leaded by complex indoor en-vironment,the paper puts forward a BP training method to replace traditional solutions,in order to eliminate the ex-cessive experience dependence and enhance algorithm＇s environmental adaptability. Based on that,the paper pro-cesses the statistical data to study the relationship between measured distance and predicated distance,and buildsthe probabilistic annular intervals of anchor nodes. On this basis,an algorithm of overlapping intervals is proposed,and finally confirm a new hybrid localization scheme with the supplementary method of multiple boundary centroidalgorithm. The scheme is tested on CC2420/Tiny OS platform to validate its adaptability and performance,and re-sults showed that the location accuracy is improved by 17.2% compared to traditional weighted centroid algorithmand 37% compared to centroid algorithm.
出处
《传感技术学报》
CAS
CSCD
北大核心
2015年第12期1823-1829,共7页
Chinese Journal of Sensors and Actuators
基金
浙江省自然科学基金项目(LY15F020010
LY12F02013)
宁波市移动网络应用技术创新团队项目(2011B81002)
浙江公益技术研究工业项目(2014C31059)
软件工程国家重点实验室开放课题项目(SKLSE2014-10-05)
嵌入式与服务计算教育部重点实验室开放课题项目(ESSCKF201302)
宁波大学研究生重点课程建设项目(ZDKC2013003)
