Research On Indoor Service Robot Localization Technology Based On Sensor/Actuator Network

Along with the number of elderly and Empty Nester increased dramatically.The demand for real-time monitoring of the elderly is also growing fast.The set of rehabilitation,elderly,nursing the old,doing housework in one of the indoor service robots are playing an increasingly important role in the social life.The localization technology is an important direction of service robot.It has important theoretical significance and practical value.In this thesis,we introduced the current research status of the indoor localization and sensor/actuator networks.Then we analyze the indoor positioning technology based on wireless sensor network,common indoor localization algorithm and localization system.As the actual experiment found that the indoor environment is too ideal in traditional positioning system and using a single model,which can lead to the problem of low positioning accuracy and non real time.For this problem,we proposed an interactive multiple model localization algorithm based on RSSI ranging.Indoor environment is very complex,when the moving target moves in indoor environment,the communicating signal between moving node and anchor nodes change between LOS and NLOS frequently and randomly,which has negative effect on the accuracy of location estimation.In this paper,the shadowing model of RSSI ranging is set up.The least squares fitting method is used to fit the ranging model and estimate model parameters of LOS and NLOS condition respectively.In this way,the LOS and NLOS ranging model is established.And then we put forward the R-IMM-RLS(RSSI-Interactive multiple model-Recursive least-squares method)localization algorithm.Two parallel EKFs with different parameters are employed in order to filter the measuring distance at the same time in IMM frame,both of the mode probabilities are calculated by the mode likelihoods and history probabilities.The mode transition between LOS and NLOS mode is based on Markov chain and mode probabilities.The weighted mean of the two modes filtering results is taken as the estimated distance of IMM.As the least square method of inversion result in low efficiency and non real-time positioning,the RLS(recursive least-squares method)is proposed to solve the process of inverse matrix.The RLS lead the system to provide accurate online real-time positioning.Based on above condition,the indoor mobile robot localization system is designed,including the coordinator,the reference node,the blind node,PC machine and detailed system positioning principle and software design.The experimental results show that the R-IMM-RLS localization error is 0.52 meters,which is o'bviously lower than the only LOS model of 0.97 meters and only NLOS model of 1.14 meters,and the speed of the recursive least squares method is 73.9%higher than that of ordinary least squares method.