The Research On Distributed Control Method Of Multi-UAV Cooperative Target Tracking

The coordinated target tracking of multi-fixed-wing UAV systems has great appli-cation value in the fields of environmental protection monitoring,border patrol,resource search and etc.However,the coordinated target tracking control of multi-fixed-wing UAV systems still exists several unsolved challenges such as the constrained control input,com-plex system model dynmics which is hard to be modeled,the difficulties to optimize col-laborative control methods and etc.This paper focuses on the cooperative target tracking control problem of multiple fixed-wing UAVs and aims to make breakthoughs in the fol-lowing problems:the optimal control design of nonlinear systems subject to asymmetric constraints;the distributed optimal control of nonlinear multi-agent system and the adap-tive attitude control of the fixed-wing UAV.This paper provides an integrated solution for the multi-fixed-wing UAV cooperative target tracking from the upper-level optimization of the navigation algorithm to the bottom-level design of the attitude adaptive control.The methods proposed in this paper effectively improve the coordinated target tracking perfor-mance of fixed-wing UAVs.The main work and innovations of the paper are summarized as follows:(1)Aiming at solving the optimal control design problem for the non-standard system with unsymmetrical input constraints,this paper proposes an online opti-mal control algorithm based on the policy iteration strategy.The theoretical results obtained in this paper break through the theoretical limitation that the existing op-timal control theory can only deal with standard systems.Meanwhile,the method proposed in this paper is applied to solve the optimal circumnavigation control prob-lem for the fixed-wing UAV.There are still two limitations in the existing optimal control study for the input-constrained systems.Firstly,the system’s dynamics are required to be in standard form,i.e.,the system should be stabilizing at the origin when the control input is zero.Secondly,the input constraint is required to be symmetrical,i.e.,the admissible interval of the control input is a symmetrical interval centered at the origin.However,in many application scenarios(such as the UAV circumnavigation tracking),the control systems are often non-standard systems with unsymmetrical constraints.In order to lay a good theoretical foundation for the study of the optimal target tracking control of UAVs,this paper deeply studies the optimal control problem of nonlinear system in a more gen-eral form.An online optimal control algorithm based on the policy iteration strategy is designed for the non-standard system with unsymmetrical input constraints.Meanwhile,the convergence and stability of the proposed algorithm are strictly proved.(2)Aiming at solving the problems of non-identical and unknown input con-straints in the multi-UAV system,a cooperative circumnavigation target tracking control algorithm based on the Lyaponov method is proposed.The proposed method can achieve the cooperative target tracking of UAVs with different maneuverabili-ties.In order to solve the problem of coordinated target tracking of multi-UAV systems with different maneuverability,this paper proposes a multi-UAV cooperative circumnav-igation target tracking control algorithm based on the Lyaponov method.The proposed algorithm can drive the UAVs constrained by different input saturations to achieve the co-operative circumnavigation target tracking without knowing the specific input constraints of each UAV.The paper strictly proves that as long as the UAVs in the system meet the minimum maneuverability requirements for circumnavigation tracking,the multi-UAV system can achieve the coordinated circumnavigation tracking of a moving target.The obtained theoretical results can be used as the initial admissible control law of the itera-tive algorithm proposed in this paper when seeking the optimal cooperative target tracking control law.(3)Aiming at solving the problem of“the curse of dimension”in the cooper-ative optimal control,a multi-UAV cooperative tracking optimal control algorithm is proposed by combining the mean field and policy iteration method.The proposed algorithm effectively reduces the data samples and time required by the training of the cooperative optimal control law.In order to solve“the curse of dimension”prob-lem in the cooperative optimal control of multiple UAVs,the paper models the input of all neighboring UAVs as a variation near the average value in a steady state with the aid of the“mean field”idea in the current multi-agent learning.Then,the cooperative cir-cumnavigation target tracking control problem is transformed into a standard single-agent robust optimal control problem.Combined with the designed optimization index based on the accumulated Fisher information,a multi-UAV optimal circumnavigation control law for target tracking is designed.The proposed training algorithm compresses the sample sampling space from the high-dimensional space of the multi-agent state coupling into a low-dimensional space of the single-agent state.The proposed algorithm reduces the training time complexity from O(m~n)to O(m),where m is the state number of the single agent and n is the agent number of the multi-agent system.(4)Aiming at solving the problem of fitting error in the optimal control algo-rithm based on policy iteration,the paper analyzes the tolerance boundary condi-tion of the fitting error,and further proposes a new optimal control algorithm which takes the fitting error into consideration.The proposed algorithm improves the fit-ting accuracy compared with the traditional fitting algorithm.There often exists fitting error when solving the optimal control strategy of a nonlinear system.The tradi-tional optimal control algorithm based on the policy iteration strategy usually ignores the influence of the fitting error on the optimization effect of the control strategy,which leads to the poor performance when it is applied to solve the optimal nonlinear control problem such as circumnavigation control.To tackle this issue,this paper deeply studies the effect of fitting error on the optimal control strategy based on policy iteration,and analyzes the fitting error tolerance boundary conditions which can ensure better performance of the control strategy after iteration.Then,by designing a sequence of positive semi-definite functions,the paper proposes an optimal control algorithm considering the fitting error by using a method based on local space fitting.Compared with the traditional fitting algo-rithm,the algorithm proposed in this paper improves the fitting accuracy by two orders of magnitude.(5)Aiming at solving the problem of the complex dynamics of fixed-wing UAV which is difficult to be accurately modeled,a yaw rate adaptive controller based on neural network feedforward control is proposed,which can realize the UAV sys-tem’s tracking to a given yaw rate with the UAV’s dynamics unknown.The proposed method effectively solves the connection issue between the upper-level navigation op-timization algorithm and the underlying adptive attitude controller.In view of the fact that the dynamic model of the attitude control layer of the fixed-wing UAV is often difficult to be accurately modeled,this paper designs a control architecture which consists of an optimal yaw rate planner and an adaptive yaw rate tracker.The optimal yaw rate planner is responsible for planning the optimal yaw angular velocity of the UAV for the specified task,and the yaw angular velocity tracker of the attitude control layer is responsi-ble for tracking the planned yaw angular velocity.The control architecture designed in this paper makes the upper-layer navigation optimization algorithm do not need to consider the dynamic model of the actual UAV attitude control layer,which reduces the difficulty of solving the optimization algorithm.Besides,the attitude controller design does not need to consider the specific task planning algorithm,and only focuses on how to improve the tracking performance.This paper designs an adaptive controller of the UAV’s yaw an-gular velocity based on neural network feedforward control.The proposed algorithm can achieve the tracking of a given yaw angular velocity with the system’s dynamics unknown.