Research On Brain-inspired Intelligence Decision And Coordination Control For Multiple Unmanned Aerial Vehicles

With the increasing complexity of environment and diversity of tasks,multiple unmanned aerial vehicles(UAVs)systems are gradually becoming an important research direction in the future.Since the current intelligence level of UAVs is difficult to meet the requirements of complex tasks,this dissertation investigates the intelligent task decision-making and cooperative control problem for multiple UAVs,in order to improve the intelligence and collaboration capabilities of multi-UAV system.The main content of the thesis is as follows:Firstly,the priority-based centralized task decision-making problem for multiple UAVs is studied.Consider computational burden of traditional method and the effect of environment uncertainty,a brain-inspired intelligent method is proposed based on four parts,including experience summarization,behavior prediction,critic&improvement,online decision-making,by mimicking the decision-making process of human brain.The three parts of them,experience summarization,critic&improvement,online decision-making are used to settle the mentioned task decision-making problem,which improving the solving speed and the ability to deal with environmental uncertainty,then the UAVs for tasks with different priority are allocated appropriately.Secondly,the distributed task decision-making problem for hunting task is investigated.Consider the effect of partially state observation,environmental dynamics and uncertainty,the problem is solved through the proposed brain-inspired intelligent method.First,according to the gained experience,the initial decision scheme and reward&punishment mechanism are summarized and the experience database is established.Then,the policy of opposite UAVs is inferred and their next action are predicted.Furthermore,based on the centralized-critic&decentralize-actor structure,the policy of UAVs is improved.Finally,the online decision-making is realized through trained network and when the decision result is not satisfying,close-loop iterative strategy is established.Thirdly,the cooperative fault tolerant control problem for multiple UAVs under collision avoidance constraint is investigated.Considering the effect of lumped disturbances,including model uncertainty and external disturbance,a novel adaptive distributed sliding mode fault tolerant controller is designed based on leader-follower structure and prescribed performance function,by taking the advantage of sliding mode method,which is robust and insensitive to disturbance,and prescribed performance method,which can quantitative describe the transient and steady states performance.The strategy effectively solves the discontinuity problem of sliding mode control signal and does not need any knowledge of the boundary of lumped disturbances.Furthermore,collision avoidance constraint is satisfied through forcing the relative position inside the prescribed performance region,which is independent of lumped disturbances and actuator fault.Finally,the task decision-making and cooperative control problem for multiple UAVs based on a competitive manner is studied.Since the complex coordination behavior in nature can be generated by competition,the coordination task that a group of agents to track a moving target is considered.The k-winners-take-all,which can describe competitive nature is used to model this coordination behavior.By taking the advantage of dual neural network,which is high parallelism and fast solution,and finite time method,which demonstrate nice features such as fast convergence,high accuracy as well as good disturbance rejection properties,a novel finite-time task decisionmaking and cooperative control method based on adaptive dual neural network and disturbance rejection is proposed.Through this method,the nearest UAVs will perform capturing tasks,and the fast decision-making and cooperative control is realized.