Research On Robust Consensus Tracking Control Of Multi-agent Systems Under Complicated Constraints

The distributed cooperation of the multi agent systems,where the agents interact with each other to enhance the system capability,can accomplish more complex tasks,and is a crucial way to to improve the efficiency of the future autonomous systems in both the military and civil fields.This dissertation studies robust distributed cooperative consensus tracking control under multiple constraints and uncertainties.Specifically,the impacts on the behavior of multi-agent systems are taken into account,including not only the actuator constraints like actuator saturation and maneuverability limitations,but also the unknown factors such as model uncertainties and external disturbances.The proposed control solution provides stability and robustness to the system,and it is further extended to the problem of robust formation control for multi-agent systems under complex conditions.The main innovative achievements are as follows:1.This dissertation studies the distributed control method based on the bounded and measurable adaptive variable input.Robustness conditions are analyzed,which reveals the effects of input saturation constraints on the motions of the multi-agent systems under consensus tracking.To convert the consensus problem into a system-state stabilization problem in an auxiliary coordinate,a saturated input model is first developed based on the prior bounded local neighborhood synchronization error(PB-LNSE),and then a coordinate transformation and a Lyapunov function with respect to the integral of the tracking error are constructed.Lagrange's median theorem is used to perform a linear transformation to the error transformation matrix to prove the equivalent relationship between the convergences of PB-LNSE and that of the tracking error.The asymptotic convergence condition on the adaptive variables is obtained,which is further verified to guarantee the consensus tracking control of the system.The consensus tracking control method is extended to address the formation tracking control problems by introducing formation variables.The formation control problem is transformed into the consensus problem by means of the coordinate transformation,based on which the convergence of the formation error is proven.Compared with the control strategy based on the variable threshold constraint,the control input signal generated by the proposed control is continuous,and therefore the signal jumping phenomenon is prevented.2.This dissertation studies the mechanism of complex uncertainty factors' affecting the consensus of multi-agent systems under the control input constraints.Under the condition of the lack of velocity measurement,we construct an uncertainty and disturbance estimator(UDE)for the robust cooperative control scheme to suppress unknown external disturbances and measurements noise.By introducing a parameter mapping in the UDE,the steady-state tracking control error is reduced via tuning a single parameter for the stabilization.The UDE is used to compensate for the complex unknown components in the system dynamics,to mitigate the influence of uncertainties on the system performance.Moreover,the generated continuous estimation signals avoid input chattering.Formation control of a multi-agent system in a complex environment is achieved by applying UDE to the controller,and the formation control error can be reduced by adjusting a single parameter.Furthermore,the consensus tracking control method for the particle double-integrator model is extended to the cluster control of quadrotor.By establishing a six-degree-of-freedom controller model,the height consensus tracking of quadrotor without velocity measurement and under lift saturation constraints is achieved,and simulation experiments are designed to verify the effectiveness of the method.3.A neural network estimation-based robust cooperative control method is studied for a multi-agent system with complex constraints such as model uncertainties and time-varying disturbances.The adverse effects of complex model uncertainties on system performance are resolved,and consensus tracking control of a second-order multi-agent system under input saturation constraints is delivered.Under the global saturation of the control input,a neural network is constructed to estimate the unknown time-varying components of the state variables in the system model.High-accuracy approximations of these unknown components are achieved by introducing the estimation error terms which are proven to be bounded.An adaptive consensus control scheme is designed,and the update law of adaptive parameters is derived based on Lyapunov stability conditions to smooth the trajectories of the agents.Then,the control input generated by the adaptive controller is limited by the saturation function within the measurable range described by the linear expression.The saturation constraint on the control input renders the control scheme applicable to uncertain systems with arbitrary continuous nonlinear unknown components in the dynamic equations.The numerical simulation results show that the model uncertainty estimation error is bounded and the controlled system is robust.Compared with state feedback control methods,the proposed adaptive control scheme can increase the convergence speeds of consensus states and reduce the system oscillation.4.The consensus problem of high-order multi-agent systems with restricted control inputs and Markov switching topological is studied.The problem of non-convex constraints is solved by introducing non-convex constraint operators and designing a consensus control protocol for the system based on the neighbor node information in a multi-agent system.Considering the randomness of the closed-loop system,an auxiliary matrix is constructed to convert the original closed-loop system into an equivalent system with the system matrix as the Metzler matrix.Then the properties of the non-negative matrix are used to prove that the high-order multi-agent systems can achieve mean square consensus under Markov switching of topological structure.The sufficient and necessary conditions for achieving consensus of all states are obtained,which is of a low conservative type;Compared with the actuator saturation using a given value,function or convex closed set constraint,non-convex constraint conditions are more generic.These theoretical results can be used to solve the cooperative tracking control problem for practical autonomous systems such as UAV clusters under unwinding conditions or unmanned ship clusters underwater conditions.Finally,numerical simulations are carried out for the third-order,fourth-order,and sixth-order multi-agent systems,to verify the effectiveness of the method.