Research On Indoor Multi-Agent Cooperative Localization Method Based On Artificial Landmarks

Currently,indoor positioning services for intelligent agents have broad application prospects.Among them,the positioning method based on visual sensors has become the mainstream method to solve the problem of indoor positioning of intelligent agents at this stage.However,when a single intelligent agent relies solely on visual information for positioning,there are often problems of insufficient accuracy and robustness.Therefore,this thesis proposes a multi-intelligent agent cooperative positioning method based on artificial landmarks,the main research content of which is as follows:(1)Estimation of intelligent agent states parameters,including angle,distance,and speed parameters.According to the structural characteristics of the designed artificial landmarks,the corresponding recognition algorithm is used to obtain angle information in the image.The YOLOv3 network is trained on a self-built data set with manual annotation to achieve intelligent target recognition and angle solution.Through the fixed threshold method,the TOA estimation of the composite HFM signal is realized,so as to obtain the distance and speed between intelligent agents.The experimental results show that the estimation error of this method is within the allowable range,and the recognition accuracy is relatively high.(2)Research on multi-intelligent agent cooperative positioning method based on particle filter.After comparing the advantages and disadvantages of extended Kalman filter and particle filter,the particle filter is selected as the information fusion algorithm.The motion models of single and multiple intelligent agents are analyzed,and the corresponding particle filters are established to achieve multi-source information fusion.The numerical simulation results show that this method has high positioned robustness and can achieve sub-meter-level positioning.(3)Based on the above research methods,an intelligent agent positioning experimental platform was constructed,and static and dynamic experiments of single intelligent agent selfpositioning and multi-intelligent agent cooperative positioning were carried out.The experimental results show that the average positioning error of this method is 0.34 m,and there is an 80% probability that it is less than 0.46 m,which verifies the effectiveness and reliability of the multi-intelligent agent cooperative positioning method proposed in this thesis.