运动行为辅助的多源数据自适应融合室内定位方法

Smartphone-based indoor localization has wide application future in many applications of localization-based service, such as pedestrian navigation, intelligent transportation and so on. The positioning accuracy of the single data source-based indoor localization is poor. Multisource data fusion is the development trends of indoor localization. Indoor environment, user, smartphone usage mode, and smartphone type are complex for indoor localization, draws high demands in adaptive ability of the multisource data fusion. Based on activity recognition, the relative trajectory can be inferred. The time series multi-source data can be associated with each other by the relative trajectory. The pedestrian’s activity in an indoor environment is restrained by the indoor map; the historical trajectory can be inferred by activity recognition. Based on the observations, this project proposes an activity assisted-multisource data self-adaptive fusion method for indoor localization. The detailed research contents are as follows: activity recognition in different usage modes of smartphone, activity-assisted multisource data fusion method for indoor localization, adaptive parameters learning for the multisource data fusion model, and experiments in typical indoor environments. This project tries to solve the following two problems: how to build the activity-based multisource data fusion model for indoor localization and how to learn the adaptive model parameters. Through the study of this project, the positioning accuracy and adaptive ability of the smartphone-based indoor localization can be improved, and this project can provide theoretical support and technical guidance for the practical applications of indoor localization.

基于智能手机的室内定位在行人导航、智能交通等位置服务领域具有广泛应用前景。依靠单源数据的室内定位方法精度较差，多源数据融合是室内定位的发展趋势。室内场景、用户、手机使用模式及手机种类的复杂性，对多源数据融合算法的自适应能力要求较高。通过运动行为识别可以推算相对运动轨迹，关联智能手机获取的时序多源数据；行人在室内的运动行为受室内地图约束，基于运动行为识别可以推算用户的历史轨迹。本项目以此为出发点，研究运动行为辅助的多源数据自适应融合室内定位方法。具体研究内容包括：多使用模式下的运动行为识别、运动行为辅助的多源数据融合室内定位、多源数据融合模型的参数自适应学习方法以及实验验证与结果分析。本项目重点突破运动行为辅助的室内定位多源数据融合模型构建及模型参数自适应学习的难题。通过本项目的研究，可以提高基于智能手机室内定位算法的精度及适用性，为室内定位的相关应用提供理论支持和技术指导。

基于智能手机的室内定位在行人导航、智能交通等位置服务领域具有广泛应用前景。依靠单源数据的室内定位方法精度较差，多源数据融合是室内定位的发展趋势。室内场景、用户、手机使用模式及手机种类的复杂性，对多源数据融合算法的自适应能力要求较高。通过运动行为识别可以推算相对运动轨迹，关联智能手机获取的时序多源数据；行人在室内的运动行为受室内地图约束，基于运动行为识别可以推算用户的历史轨迹。本项目以此为出发点，研究运动行为辅助的多源数据自适应融合室内定位方法。具体研究内容包括：多使用模式下的运动行为识别、运动行为辅助的多源数据融合室内定位、多源数据融合模型的参数自适应学习方法以及实验验证与结果分析。本项目重点突破运动行为辅助的室内定位多源数据融合模型构建及模型参数自适应学习的难题。通过本项目的研究，可以提高基于智能手机室内定位算法的精度及适用性，为室内定位的相关应用提供理论支持和技术指导。