Research On Key Technology Of Pedestrain Autonomous Location Based On Inertial Sensors

With the rapid development of society and the continuous progress of science and technology,there is an increasing demand for accurate positioning and navigation in daily life,national defense and other fields.However,because the satellite signal can't penetrate most obstacles and there is multipath effect,the global positioning system can't be used effectively in indoor,underground,tunnel and forest,and can't meet the needs of people's daily life.The Inertial Navigation System(INS)doesn't need to rely on external signals,but only depends on the inertial signals measured by its own inertial sensors to achieve autonomous navigation,which overcomes the above disadvantages of satellite navigation and has always been one of the research hotspots in the field of positioning and navigation.Based on inertial navigation technology,this paper explores the key technologies of pedestrian autonomous location.In this paper,the inertial sensors used in pedestrian inertial navigation and location are introduced in detail at first,and the errors of inertial sensors are analyzed.Then it introduces the relevant theory of coordinate system and the related solution of inertial navigation system,and focuses on the rotation of coordinate system and attitude matrix.Subsequently,the principle of Kalman Filter(KF)is introduced and derived in detail,including kalman filter suitable for linear systems and extended kalman filter suitable for nonlinear systems.Because the pedestrian inertial navigation system is a non-linear system,this paper also deduces the inertial navigation algorithm based on the Extended kalman filter.Because the pedestrian inertial navigation system mainly uses the double integration of the accelerometer to obtain the position information,it may bring very serious cumulative error and lead to the rapid divergence of the pedestrian track.In view of this,the classical zero-speed detection scheme in pedestrian inertial navigation system is studied in detail in this paper.it includes zero-speed detection algorithm based on fixed threshold,zero-speed detection algorithm based on variance,zero-speed detection algorithm based on difference and zero-speed detection algorithm based on generalized likelihood ratio test.These algorithms are based on a certain threshold,and their robustness is not high.In view of the low robustness of traditional classical solutions,this paper proposes a zerospeed detection algorithm based on neural network,which can accurate judgments to a variety of motion types and greatly enhance the generalization performance.At the same time,the precision,recall and F1-score of average score of Zero Velocity Update(ZUPT)points and non-Zero Velocity Update points can reach more than 99%.Finally,based on the proposed zero-speed detection algorithm based on neural network,a real-time Pedestrian Dead Reckoning(PDR)based on SoC is realized.The terminal is portable and can easily draw the pedestrian positioning trajectory in real time,and in terms of energy conversion efficiency,the conversion efficiency of FPGA is 12.58 times and 5.42 times higher than that of CPU platform and GPU platform,respectively.After a large number of experimental analysis,the average error rate of the closed-loop trajectory of the experiment on the terminal is only 1.1%.