Research On Cooperative Navigation Algorithm Of Master-slave UAV Formation

With the emergence of cooperative tasks in recent years,a single unmanned aircraft vehicle(UAV)is not enough to meet all the needs,and relying on its own individual navigation to execute tasks has become unreliable.Meanwhile,UAV formation can better complete complex operations at current stage.A high precision navigation system is a prerequisite for the efficient and successful completion of UAV formation operations.Currently,GNSS/INS combined navigation is a common navigation method,but various spatial interferences can lead to reduced navigation accuracy in complex environments where GNSS signals are limited and the statistical characteristics of noise are changed in practical applications.The use of relative information between carriers to achieve cooperative navigation has become the mainstream navigation solution for clusters.At present,cooperative navigation methods with parallel formation structure are too costly and difficult to practice,while cooperative navigation with master-slave structure has received much attention due to its wider applicability and is the focus of current research.In order to improve the navigation accuracy of master-slave UAV formations,this paper presents an in-depth study of cooperative navigation algorithms in complex environments.The main work is as follows:An information fusion algorithm based on improved particle filter is proposed for a singlemaster-multi-slave formation structure.Firstly,several optimal importance sampling and resampling methods and their advantages and disadvantages are described in detail.Secondly,the observation model is established with the high-precision INS and GPS navigation information from the master,combined with the low-precision sensors on the slaves.In order to solve the importance density function selection and particle degradation problems of particle filter,the Levenberg-Marquardt iterative method is introduced to ensure the stability and convergence of the filtering on the basis of extended particle filter;a fast resampling strategy is proposed in the resampling stage to classify the obtained particle set,with no more resampling for medium weight particles and the rest of the particles in the normalization process The adaptive weight factor optimization is used to make the obtained sample particle weights more uniform and reduce the computational complexity.Finally,it is verified through comparative experiments that the algorithm can effectively improve the localization accuracy and computational efficiency in complex environments.To address the situation that the navigation accuracy is degraded due to the failure of the single master mode host,this paper a multi master and multi slave hierarchical formation configuration,and proposes a distributed data fusion method based on the factor graph theory,which can solve the problem of poor real-time performance caused by too much information when each slave interacts with multiple hosts in the adjacent range in a dense cluster.Based on the communication distance between the master and slave drones,a maximum correntropy factor graph is constructed using correlation entropy as measurement information,and message passing algorithms are combined to achieve information exchange.The optimal estimation is solved based on the maximum correntropy criterion and measurement information.At the same time,an adaptive kernel width selection method based on observation error is designed to further improve the stability of the algorithm.Through simulation comparison with other algorithms,this algorithm has robustness,effectiveness,and reliability under abnormal noise.