基于识别图模型的多足机器人导航数据提取信息方法  

Information Extraction Method of Beidou Navigation Data Based on Identification Graphical Model

作　　者：刘罡[1] 朱茂飞 张春鹏[1] 钟华 LIU Gang;ZHU Maofei;ZHANG Chunpeng;ZHONG Hua(School of Advanced Manufacturing Engineering,Hefei University,Hefei 230601,China)

机构地区：[1]合肥学院先进制造工程学院,安徽合肥230601

出　　处：《佳木斯大学学报（自然科学版）》2022年第1期45-48,150,共5页Journal of Jiamusi University：Natural Science Edition

基　　金：合肥学院科学研究发展基金项目(20ZR01ZDA);新能源汽车产业创新发展和推广应用政策支持研发创新项目(wfgcyh2020447)。

摘　　要：从传感器数据中学习行为模式对于高级活动推理极其重要。展示了如何从北斗导航数据的痕迹中提取多足机器人的活动和重要的地方。设计的系统使用识别图模型中的分层条件随机场来生成多足机器人的活动和地点的一致模型。实验显示了系统能够使用收集的数据中训练的模型来稳健地估计多足机器人的活动。Learning behavioral patterns from sensor data is extremely important for advanced activity reasoning.This paper shows how to extract the movements and important places of the multi-legged robot from the traces of Beidou navigation data.The designed system uses a hierarchical conditional random field to generate a consistent model of the movement and location of the multi-legged robot.Experiments show that the system is able to use the models trained in the collected data to estimate robustly the movements of the multi-legged robot.

关 键 词：识别图 条件随机场 多足机器人 北斗导航数据 

分 类 号：TN964[电子电信—信号与信息处理]