arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2005.07539eess.SP
收藏

使用智能手机传感器的高级自感知导航的上下文检测

Context Detection for Advanced Self-Aware Navigation using Smartphone Sensors

Han Gao, Paul D. Groves

摘要:

导航和定位系统取决于操作环境和主车辆或用户的行为。环境决定了可用于定位的无线电信号的类型和质量,而行为可以为导航解决方案提供额外的信息。为了跨不同的上下文进行操作,上下文自适应导航解决方案通过检测操作上下文并相应地配置定位系统来引入自我意识的元素。本文提出了环境和行为上下文作为一个整体的检测,建立上下文自适应导航系统的基础。通过监督式机器学习算法,使用来自加速度计、陀螺仪、磁力计和气压计的测量结果对行为背景进行分类,总体分类准确率为95%。一个连接相关的过滤器,然后实施,以改善行为检测结果。从GNSS测量中检测环境背景。使用概率支持向量机(SVM)将它们分为室内,中间和室外类别,然后使用隐马尔可夫模型(HMM)进行时域过滤。由于永远不会有完全可靠的上下文检测,本文还展示了环境和行为关联如何有助于减少上下文确定算法选择不正确上下文的机会。最后,提出的上下文确定算法进行了测试,在一系列的多上下文场景。

英文摘要:

Navigation and positioning systems dependent on both the operating environment and the behaviour of the host vehicle or user. The environment determines the type and quality of radio signals available for positioning and the behaviour can contribute additional information to the navigation solution. In order to operate across different contexts, a context-adaptive navigation solution introduces an element of self-awareness by detecting the operating context and configuring the positioning system accordingly. This paper presents the detection of both environmental and behavioural contexts as a whole, building the foundation of a context-adaptive navigation system. Behavioural contexts are classified using measurements from accelerometers, gyroscopes, magnetometers and the barometer by supervised machine learning algorithms, yielding an overall 95% classification accuracy. A connectivity dependent filter is then implemented to improve the behavioural detection results. Environmental contexts are detected from GNSS measurements. They are classified into indoor, intermediate and outdoor categories using a probabilistic support vector machine (SVM), followed by a hidden Markov model (HMM) used for time-domain filtering. As there will never be completely reliable context detection, the paper also shows how environment and behaviour association can contribute to reducing the chances of the context determination algorithms selecting an incorrect context. Finally, the proposed context-determination algorithms are tested in a series of multi-context scenarios.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑