IMU Indoor Positioning Algorithm Based On Improved Dead Reckoning And Particle Filter

With the rapid development of the Internet of things and the rise of the wisdom city construction, the demand of indoor localization services such as target discovery, medical services, and smart home becomes increasing. The outdoor positioning accuracy of traditional GPS and cellular network technology is better, but due to the block of wall indoor, radio-frequency signal is intervened and positioning accuracy is greatly lower.Indoor positioning technology at present is mainly based on a variety of wireless network technology, such as Wi-Fi, RFID(Radio Frequency Identification), WLAN(Wireless Local Area Networks), etc., which using the RSS(received signal strength) to realize the indoor positioning. But because of multipath effect, signal fading and walking around,indoor environment becomes more complex and change of RSS is various than usual.The indoor positioning accuracy based on wireless network is restricted, which needs auxiliary signal base station. Based on the above reasons, this thesis proposes IMU(Inertial Measurement Unit) positioning, in which using Inertial Measurement Unit to calculate the user’s location at the next moment and estimate the walking path. IMU positioning does not need auxiliary signal base station, which with full autonomy and not affected by the external environment. Inertial Measurement Unit has been widely installed in the intelligent mobile devices, and doesn’t need to add new components in application. This thesis proposes IMU indoor positioning algorithm based on improved dead reckoning and particle filter. In this thesis, the main research contents and innovation points include:1. It is proposed indoor localization algorithm framework based on IMU, which including steps detection, step length estimation and particle filter algorithm. In view of the great error in the data processing and the low positioning accuracy of the original dead reckoning, data processing is optimized on the original model which puts forward the step detection algorithm and adaptive step length estimation algorithm. By introducing time window detection and dynamic time warping algorithm, step detection algorithm limits the time interval of steps and compares the similarity of adjacent step toeliminate the error data, making the detection more accurate. Using the relationship between the step length and step frequency, acceleration and linear regression algorithm,adaptive step length estimation algorithm makes close to actual step length, reducing the step length estimation error.2. Particle filter is introducing on the basis of dead reckoning, accomplishing location information integration and optimization process. It is one of the most important aspects of the localization algorithm. The improved particle filter based on the geometric center and likelihood estimation is proposed in order to solve the problem of particle dilution and degradation. In re-sampling phase, using geometrical center to implement particle re-sampling, it filters then re-sampling particle by the distance between the particle and geometric center. In the phase of particle weighting calculation, in view of the non-stationary non-Gaussian noise results in low accuracy of state estimation and divergent tend problem of the particle filter algorithm, it adopts non-Gaussian noise parameter estimation based on likelihood estimation to approximately estimate the measurement noise, instead of the Gaussian density function. It not only solves the problem of particle dilution and degradation, but also improves the positioning accuracy.It has been tested in simulated environment by Matlab2010 a including step detection,step length estimation, improved particle filter and IMU localization algorithm.Performance analysis and simulation results show that proposed algorithm is reasonable and effective. And the algorithm can improve the positioning accuracy and reduce the positioning error.