基于智能手机的室内定位在研究和工业领域都引起了相当大的关注。然而在复杂的定位环境中,定位的准确性和鲁棒性仍然是具有挑战性的问题。考虑到行人航位推算(PDR,pedestrian dead reckoning)算法被广泛配备在最近的智能手机上,提出了...基于智能手机的室内定位在研究和工业领域都引起了相当大的关注。然而在复杂的定位环境中,定位的准确性和鲁棒性仍然是具有挑战性的问题。考虑到行人航位推算(PDR,pedestrian dead reckoning)算法被广泛配备在最近的智能手机上,提出了一种基于双延迟深度确定性策略梯度(TD3,twin delayed deep deterministic policy gradient)的室内定位融合方法,该方法集成了Wi-Fi信息和PDR数据,将PDR的定位过程建模为马尔可夫过程并引入了智能体的连续动作空间。最后,与3个最先进的深度Q网络(DQN,deep Q network)室内定位方法进行实验。实验结果表明,该方法能够显著减少定位误差,提高定位准确性。展开更多
In this paper, we integrate inertial navigation system (INS) with wireless sensor network (WSN) to enhance the accuracy of indoor localization. Inertial measurement unit (IMU), the core of the INS, measures the accele...In this paper, we integrate inertial navigation system (INS) with wireless sensor network (WSN) to enhance the accuracy of indoor localization. Inertial measurement unit (IMU), the core of the INS, measures the accelerated and angular rotated speed of moving objects. Meanwhile, the ranges from the object to beacons, which are sensor nodes with known coordinates, are collected by time of arrival (ToA) approach. These messages are simultaneously collected and transmitted to the terminal. At the terminal, we set up the state transition models and observation models. According to them, several recursive Bayesian algorithms are applied to producing position estimations. As shown in the experiments, all of three algorithms do not require constant moving speed and perform better than standalone ToA system or standalone IMU system. And within them, two algorithms can be applied for the tracking on any path which is not restricted by the requirement that the trajectory between the positions at two consecutive time steps is a straight line.展开更多
文摘基于智能手机的室内定位在研究和工业领域都引起了相当大的关注。然而在复杂的定位环境中,定位的准确性和鲁棒性仍然是具有挑战性的问题。考虑到行人航位推算(PDR,pedestrian dead reckoning)算法被广泛配备在最近的智能手机上,提出了一种基于双延迟深度确定性策略梯度(TD3,twin delayed deep deterministic policy gradient)的室内定位融合方法,该方法集成了Wi-Fi信息和PDR数据,将PDR的定位过程建模为马尔可夫过程并引入了智能体的连续动作空间。最后,与3个最先进的深度Q网络(DQN,deep Q network)室内定位方法进行实验。实验结果表明,该方法能够显著减少定位误差,提高定位准确性。
基金Project(61301181) supported by the National Natural Science Foundation of China
文摘In this paper, we integrate inertial navigation system (INS) with wireless sensor network (WSN) to enhance the accuracy of indoor localization. Inertial measurement unit (IMU), the core of the INS, measures the accelerated and angular rotated speed of moving objects. Meanwhile, the ranges from the object to beacons, which are sensor nodes with known coordinates, are collected by time of arrival (ToA) approach. These messages are simultaneously collected and transmitted to the terminal. At the terminal, we set up the state transition models and observation models. According to them, several recursive Bayesian algorithms are applied to producing position estimations. As shown in the experiments, all of three algorithms do not require constant moving speed and perform better than standalone ToA system or standalone IMU system. And within them, two algorithms can be applied for the tracking on any path which is not restricted by the requirement that the trajectory between the positions at two consecutive time steps is a straight line.
