Design And Implementation Of Adaptive Particle Filter Reinforcement Learning Location System Based On Environment Perception

In order to comply with the trend of urban intelligence and meet the needs of users,indoor positioning technology has been greatly improved in recent years.But at present,there is much room for improvement in both algorithm accuracy and system completeness.Among them,inertial navigation and positioning technology is often affected by iron cabinets and terrain changes,and magnetic field distortion has an impact on the accuracy of direction estimation.If a method can be found to integrate the environmental information into the particle filter algorithm and endow the inertial navigation with the ability to perceive the environment,it will greatly improve the user experience.At the same time,most indoor positioning systems currently lack the ability to manage and supervise the training process of the positioning model.This also greatly increases the maintenance cost and time cost of the system,hinders the development of indoor positioning technology,and reduces the availability of the system.This topic studies the above problems.First,according to the characteristics of reinforcement learning to improve performance through self-iteration,a Particle Filter Reinforcement algorithm is proposed.By observing the distribution state of particles and local indoor maps,learning selection appropriate actions are used to control the standard deviation of the direction of particle motion;secondly,based on the needs,expand the function of the positioning platform,design and implement the model training subsystem,provide model file management and model training task management and other functions,increase the controllability of the model training process,and reduce the operation and maintenance cost and threshold of the positioning system;finally,the existing platform is optimized to realize the Particle Filter Reinforcement positioning subsystem,so that the system can play a better positioning effect in complex scenarios.Based on the software project development process,this paper first analyzes the problems existing in the existing inertial navigation and positioning technology,gives solutions,and introduces the required theory and technology;then gives the analysis results of the system’s needs,and provides a design idea for the system.From the whole to the part,the introduction is carried out,the final system design scheme is clarified,and the algorithm structure is discussed at the same time.Finally,the completed test results of the system are given to ensure that the system has no obvious defects.The research results of the positioning algorithm in this topic have been published in the IEEE journal.Compared with the traditional Particle Filter algorithm,the proposed Particle Filter Reinforcement algorithm has an accuracy improvement of more than 20%;and the development and testing of the positioning system.It has all been completed and successfully passed the project review of China Unicom Smart City Research Institute.