Target Tracking And Allocation Method For Multi-to-multi Attack-defense Game Of Exoatmospheric Interceptor

Based on the background of multi-to-multi-attack and defense game problem outside the atmosphere,this paper focuses on the key technology research of multi-interceptor game maneuvering target problem,including high-precision attitude maneuver control method considering interceptor elasticity and model parameter uncertainty,multiinterceptor cooperative tracking method for multiple targets based on passive infrared information,and multi-interceptor optimal allocation method for multiple targets based on reinforcement learning.Firstly,according to the actual situation,a typical multi-to-many attack-defense game scene outside the atmosphere is constructed,and the relevant coordinate system,dynamics and kinematics model are established,which lays a foundation for the subsequent research on interceptor attitude control method,target tracking and allocation method.Aiming at the problem of high-precision attitude maneuver control of interceptor,an adaptive attitude control method of interceptor based on multi-mode input shaping is proposed.Firstly,aiming at the problem of elastic vibration excited by the attitude and orbit control engine,the rigid-flexible coupling attitude dynamics model of the interceptor is derived.In order to avoid the decrease of maneuvering efficiency caused by attitude coupling,an attitude maneuvering baseline trajectory design method based on Euler axis angle concept is designed.Then,the multi-mode ZVD input shaping method is used to tune the attitude maneuver baseline trajectory,and the attitude maneuver reference trajectory that can effectively suppress the elastic vibration can be obtained.Finally,aiming at the residual vibration and modeling uncertainty,an adaptive feedback tracking controller based on sliding mode surface is designed to realize the stable tracking of attitude maneuver reference trajectory.In view of the multi-target tracking problem,a basic target tracking model is first established.Aiming at the heavy-tailed noise characteristics of infrared sensors,the estimation accuracy of traditional Kalman filter method is reduced.The robust cubature filter method based on student t distribution is studied to improve the estimation accuracy.Aiming at the problems of uncertain target maneuvering time,various maneuvering types and large maneuvering intensity,the research on directed graph variable structure multimodel filtering method is carried out to improve the adaptive ability of model set and reduce model competition.Combined with the above research,the directed graph multimodel multi-target robust tracking method is studied,and the corresponding simulation verification is carried out.The obtained high-precision target tracking results can provide a basis for subsequent target allocation.Aiming at the multi-objective optimal allocation problem,a multi-objective allocation model is established firstly,including interception probability and allocation efficiency.For multi-interceptors with centralized information processing,the interception allocation agent is trained based on the Munchausen-Implicit Quantile Nets(M-IQN)optimized by the Munchausen framework.For the multi-modeler with distributed information processing,considering the constraints of limited observation and limited communication between interceptors,the interception allocation agent is trained based on Optimistic Weighted-QMIX(OW-QMIX).Finally,mathematical simulation is carried out to verify and analyze the effectiveness of the above reinforcement learning algorithm for multitarget interception task allocation...