Research On Indoor Positioning Algorithm Based On TOA/PF In NLOS Environment

The 21 st century has entered the third technological revolution,various space detection technologies have also flourished.People are not only satisfied with the outdoor location services provided by satellite positioning,but also turn their attention to indoor positioning.The introduction of many new concepts such as Metaverse,AR,and VR has once again contributed to the development of indoor positioning.However,the indoor environment is relatively complex,,which seriously interferes with the signal and has various signal modes.In most cases,the communication state between the tag and the base station is non-line-of-sight(NLOS).Among them,the factor that has the greatest impact on the indoor positioning effect is the non-line-of-sight measurement value.At present,many studies have put forward improvement schemes for positioning problem in the non-line-of-sight environment.but there is little work on the identification of non-line-of-sight error sources.Therefore,this paper will start from the positioning scheme based on TOA and particle filter,combined with differential evolution algorithm and non-line-of-sight error identification strategy,to achieve high-precision indoor positioning in non-line-of-sight environment.Firstly,for the problem of insufficient indoor positioning accuracy of the least squares method in Line of Sight(LOS)positioning,we proposed a enhance opposition learning cuckoo search,(EOCS)to improve it.The opposition learning strategy compresses the search space of the algorithm,and modifies the selection mechanism of the standard cuckoo algorithm to improve the convergence speed and accuracy of the algorithm.Simulation experiments show that,under the same measurement accuracy,the improved algorithm improves the positioning accuracy by about 25% compared with the least squares method.Under the same positioning accuracy,the average position solving time is reduced by 69%compared with the standard cuckoo algorithm.Secondly,in order to achieve accurate position estimation in NLOS environment,a distributed particle filter algorithm is introduced.However,the standard filter has the problem of particle impoverishment during normal positioning,and the positioning error gradually increases with time.In view of this,this paper considers using the positioning value of the EOCS algorithm as a reference value in the normal positioning,corrects the estimated value of the particle filter that deviates from the correct likelihood region,and resets the ill-conditioned particle set by the differential optimization algorithm,and proposes a differential particle Filtering Algorithm(DEPF).The simulation results show that the DEPF algorithm can achieve higher accuracy than the standard particle filter under different measurement equations,and is more robust under the same particle number and measurement noise,which allows the DEPF algorithm to use less solution time.achieve sufficient accuracy requirements.Then,combined with the distributed particle filter strategy and the DEPF algorithm,a distributed differential evolution particle filter(DDEPF)algorithm is proposed to improve the particle depletion problem in the distributed particle filter algorithm.At this point,a relatively complete localization algorithm in the NLOS environment is obtained.Experiments show that the DDEPF algorithm can achieve effective localization when the target is in a single NLOS state for a long time.However,in the double NLOS state,the positioning performance of DDEPF is more effective,and even partial divergence occurs.For this problem,we analyzed the underlying reasons for the failure of the DDEPF algorithm in the double NLOS state,and extended the structure of the DDEPF algorithm by adding a secondary optimization process.It is set that the secondary optimization process will not be enabled when the target is in a single NLOS state,but only when it is determined that the target is in a double NLOS state.Simulation experiments show that the improved DDEPF algorithm can achieve effective tracking when the target is in a single NLOS state,and the positioning performance is no different from that of the DDEPF algorithm.When the target is in the double NLOS state,the target state can be correctly identified,and the NLOS error can be distinguished and divided,which solves the problem that the DDEPF cannot locate the target in the double NLOS state.