Research On Underground Personnel Positioning System Based On MEMS Inertial Sensor

The traditional underground positioning technology generally uses the wireless signal transmission to calculate the position.However,affected by the special environment of dust,noise and roadway,the multipath effects are serious in the underground communication,which greatly interferes with the transmission and reception of radio frequency information.As a result,the accuracy of traditional positioning technology is not high enough to meet the demand.Aiming at this problem,a method based on MEMS inertial sensor and WiFi technology is designed to improve the positioning accuracy.In this paper,at the beginning,the structure and principle of personnel positioning system based on MEMS inertial sensor is proposed.Then,the localization algorithm is described in detail,including the method of initializing the starting point and improved PDR algorithm.At last,the localization results are corrected by particle filter with cartographic information.In order to solving location error problem,the error analysis model of accelerometer,gyroscope and magnetometer is established,and the error is divided into zero value offset error and random drift error.The former compensation is simple,but the latter is the main reason for the accuracy problem of positioning.Therefore,Kalman filter algorithm is used to compensate accelerometer and improvement heuristic drift reduction algorithm is presented for gyro compensation.Meanwhile,an elliptic fitting algorithm is used to compensate magnetometer.The simulation results show that the noise of the accelerometer is significantly reduced.The random drift error of the gyroscope is improved,and the convergence time of the compensation algorithm is shortened.The validity of compensation result is also verified by Allan variance analysis.Aiming at the problem of large error in the traditional step estimation algorithm,a new method based on BP neural network has been proposed.The traditional step length estimation algorithm approximates the step length with the walking frequency and the variance of the acceleration as a linear relationship.Then a linear function is used to estimate the step length.However,in practice,there is a complex nonlinear relationship between step length and walking frequency and acceleration variance,so this method is not exact.Therefore,in this paper,the BP neural network is used to describe this nonlinear relationship.The walking frequency and acceleration variance are selected as the input value of the neural network to obtain estimated step length.Experimental results show that the proposed algorithm is about 12% higher than the traditional algorithm.In addition,we further study the effect of pedestrian height and gender for step length estimation,and put them as an additional input parameter of BP neural network,the simulation results show that after considering this two parameters,the step length estimation are closer to the real value,and the estimation accuracy is about 31.5% higher than that of the traditional algorithm.Finally,the designs of the hardware and software used in the positioning system have been introduced by this paper.Through the experimental test,the validity of the underground personnel positioning system based on MEMS inertial sensor is verified.Positioning error is about 1m when walking 100 m.