Analysis And Distributed Coordinated Control For Multi-robot Systems

As a new engine for boosting the development of national economy,the research of robotics is important indicator to measure the innovation development of the hign-end manufacturing and advanced control theory.It is of great significance to develop robotics for promoting the modern industrial system and ensuring the leading position of manufacturing.However,due to the constraints of sensors,processors,actuators and environments,the single-robot system usually has limited capabilities of perceptions,communications,calculations and executions,such that it can only be used to realize basic tasks and is always powerless for handling complex ones.With the inspiration of biological swarm behaviors,as well as with the rapid development of sensing measurements and wireless communications,multi-robot systems,possessing properties of high efficiency,low cost and strong flexibility,show great superiorities in terms of the timeliness,economy and functionality,especially for its prominent role of providing industrial productions and social services in the current normalization of the epidemic.Based on joint-space and task-space control,this dissertation respectively investigates the coordinated problems of Euler-Lagrange multi-robot systems via distributed control algorithms.The research is carried out from easy to deep,and the specific contents are as follows:This dissertation studies the formation-containment control for multi-robot systems with two-layer leaders subject to parametric uncertainties,input disturbances and directed interaction topologies.To cope with the aforementioned issues,we establish a novel formation-containment control framework,where the analysis of the systems is carried out step by step.A hierarchical controller-estimator algorithm,containing distributed sliding-mode estimators in each sub-algorithm,is proposed for the system with two-layer leaders.Moreover,by invoking finite-time stable and input-to-state stable theories,we achieve the sufficient criteria for convergence of the obtained hierarchical controllerestimator algorithm.This research provides a suitable method for solving the control problem with different model accuracy,and also provides a new idea to analyze and address the problem of multi-task control of multi-robot systems.In the presence of uncertainties,disturbances and scarce control resources,this dissertation investigates the single-and multi-synchronization control problems of networked Euler-Lagrange systems in a unified framework.Several novel event-triggered control algorithms are developed without requiring relative velocity information,which is capable of significantly mitigating the cost of unnecessary controller updates,signal transmission and computation,while possessing satisfactory control performances.Additionally,based on the Lyapunov stability techniques,the rigorous sufficient criteria for the asymptotic convergence of the synchronization errors are established,and the positive lower bounds of execution intervals are derived to exclude Zeno behaviors.Finally,by further proposing other comparative control algorithms and conceiving a performance index named trigger rate,some examples are performed to verify that the presented methods can reduce the energy consumptions of communications,calculations and executions without obvious differences of convergence performances.In this dissertation,the consensus of networked underactuated robotic systems subject to fixed and switched communication networks is discussed by developing some novel event-triggered control algorithms,which can synchronously guarantee the convergence of the active states,the boundedness of the velocities of passive actuators,and the exclusion of Zeno behaviors.In the cases of fixed networks,the sufficient criteria are established for the presented distributed event-triggered mechanisms with and without using neighbors' velocities,in order to achieve a better tradeoff between the communication load and system performance.Besides,in the situation of switched networks,the sufficient criterion is established by assuming that the union of the network has a spanning tree.A distributed sampled-data rule is constructed to decide when to update its own and neighbors' estimated positions,and thus further reduces the unnecessary control cost.Finally,by further extending the main results to three other sampled-data control algorithms,several examples with performance comparisons are provided to validate that,to some extent,the distributed/event-triggered control is more efficient than centralized/time-triggered control,and it costs the least energies for the distributed sampled-data-based event-triggered algorithm.With kinematic and dynamic uncertainties,the task-space tracking issue for multiple robotic systems is investigated in this dissertation.Two classes of control schemes are developed to address the aforementioned problem with nonredundant and redundant kinematics and input disturbances.Particularly,the property of separating of kinematic and dynamic loops is obtained by the proposed controllers,which include a control law,a kinematic parameter adaptive law and a dynamic parameter adaptive law,respectively.Moreover,by using passivity approaches,the system is proved to be passive with the external forces,applied on the end effectors,be the input signals.Consequently,the tracking errors asymptotically converge to the origin under interaction constraints,i.e.,the graph is directed,and only a subset of slave robots can access the information of the master robot.Finally,a numerical example is performed to prove that the proposed algorithm can achieve better tracking performance(i.e.,the smaller tracking errors of positions and velocities).The synchronization of multiple heterogeneous robotic systems is investigated in this dissertation with kinematic uncertainties,dynamic uncertainties and communication delays.We develop two classes of adaptive sliding-mode controllers to deal with the aforementioned problem based on directed graphs containing a directed spanning tree,which relaxes the constraints on the interaction topologies.Furthermore,by invoking techniques of Lyapunov and input-output stability theories,we obtain the sufficient conditions on asymptotic stability of the robotic systems.Thus,the position and velocity synchronization errors can asymptotically converge to the origin.By involving external disturbances,kinematic and dynamic uncertainties,this dissertation studies task-space formation tracking problem of nonlinear heterogeneous robotic systems,where the cases with both single and multiple time-varying leaders are considered.To solve the aforementioned nonlinear control problem,several novel fully distributed control algorithms,in which no global information is employed,are developed for the nonlinear systems under directed communication topologies.Based on the proposed control algorithms,the control process is classified into two parts,namely the task-space formation tracking of master robots with a single leader and that of slave robots with multiple leaders.By invoking Barbalat's lemma and input-to-state stability theory,the sufficient criteria for the asymptotic convergence of the task-space formation tracking errors are established.In addition,the obtained results are extended to formationcontainment and consensus problems in similar nonlinear cases.