Configuration Control And Planning For Multi-Agent System In Denied Environments

Multi-agent system(or MAS)is a complex network composed of multiple single and intelligent units with certain autonomous capabilities.Due to the advantages of low cost,strong robustness,and high efficiency,MASs are currently gradually replacing humans,performing dangerous or complex tasks that a single agent cannot complete in extremely demanding environments.The denial environment is a type of harsh environment where the navigation and communication are restricted and denied.It is a summary of the envi-ronmental characteristics of typical missions such as military confrontation,outer space exploration,and deep-sea research.Those harsh external conditions in the environment put forward higher requirements for the collaborative process within the MAS.In the de-nied environment,the collaboration between agents must be strengthened to promote the emergence of more functional collaborations within the MAS.Because the cooperative control and planning of configuration is an important approach to realize the emergence of MAS,it is necessary to conduct further research on the cooperative control and planning methods of the MAS configuration in the denied environment.Motivated by the necessities of MAS application in denied environment from real project,this dissertation further investigates two types of configuration control methods and two types of configuration planning methods for the MAS regarding the characteristics of the denied environment.The main research contents of this dissertation are as follows:Research on the surrounding control problem of MAS using only relative measure-ments in the navigation denied environment.The surrounding formation plays an impor-tant role in the cooperative observation,tracking,positioning and other tasks of the MAS.The expected surrounding configuration is defined as a convex hull,which is centered at the leader’s geometrical center and has a radius bigger than the maximum distance be-tween the leaders and geometrical center; Then the surrounding problem is investigated in the one-dimensional motion space considering second-order followers.A novel dis-tributed acceleration control laws are proposed using absolute positions of followers.Then when the followers cannot obtain their absolute positions in the navigation denied envi-ronment,a self-organizing algorithm is designed for the estimation of the average distance between every follower and the leader’s center based on relative measurements that are locally available in every follower.The final control protocol is derived by combining the acceleration control law with the estimation algorithm.Both theoretical analysis and simulation results validate the correctness and effectiveness.Research on the control problem of the Position Dilution of Precision(or PDOP)positioning configuration.There is a coupling effect between the cooperative localiza-tion(or CL)and the MAS configuration,which motivates the improvement of CL per-formance through configuration control.For the mobile positioning system composed of multiple agents,the PDOP distribution over target area is selected as the objective of configuration control to measure the positioning performance.Then the PDOP position-ing configuration control problem is established.An improved PDOP coverage index is designed to solve the problem of discontinuous distribution of traditional PDOP caused by the singular MAS configurations.According to the coverage control framework,the analytical partial derivative of coverage objective over every single agent state is derived and a novel distributed gradient-descent configuration control law is proposed.Finally,numerical simulations verify the correctness and effectiveness of the proposed methods,showing that it can greatly expand the coverage area of positioning service by leveraging the mobility and coordination of MAS,and is robust to node loss/resurgenceResearch on configuration planning problem based on continuous belief space.A variety of uncertainties exist in the denied environments due to both the noisy motion process and imperfect measurements.However,those uncertainties are not included in the configuration planning methods using deterministic states.Therefore,the solution generated from traditional methods may deviate from the expected performance due to the environmental disturbance during execution.Aiming at the above problems,a con-figuration planning framework using Gaussian parameterized belief space is established.The gradient of objective function is computed based on numerical difference,whereby a gradient-descent path planning in continuous belief space is proposed,which avoids the exponential complexity from the discretized methods.Finally,the numerical simulation verifies the effectiveness of the proposed method.Theoretical analysis and experimental results show that the proposed method can suppress the impact of environmental uncer-tainties and improve the robustness of planning methods.Research on configuration planning methods based on probabilistic topology.Aim-ing at the configuration planning problem where the observation sensor has a measurement boundary and the node state is uncertain,based on the previous belief space configuration planning method,we further release the usage of maximum likelihood observation for the prediction of future system behaviors within the planning horizon.The key of planning under probabilistic topology is to predict the probability of any observation connection for the disk model of sensors.Under Gaussian assumption,the distribution of a measurement connection can be modeled as the quadratic random variable in normal distribution and its probability can be computed by the finite-term approximation approach of the series expansion theorem.The truncation error of finite-term approximation is analyzed in both theoretical and experimental aspects.And a set of sufficient conditions for the stability of truncation error is summarized from these analyses.Finally,this method is applied to the configuration planning,and the configuration planning method under probabilis-tic topology is proposed.While the previous method only evaluates the future branch with maximum likelihood observation,the proposed planning method under probabilistic topology evaluates all possible branches and computes the expectation of objectives along all branches weighted by their corresponding probabilities.Therefore,the performance of configuration planning can be improved.