Research On Task Assignment Algorithms In Unmanned Swarm

Unmanned platforms are gradually replacing mankind towards the forefront of fu-ture warfare,becoming a new source of fighting capacity.As one of the main application modes of unmanned platforms,unmanned swarm has won the full attention and vigorous development of various countries with its system advantages and war efficiency.Within the unmanned swarm,a large number of heterogeneous unmanned platforms are driven by control instructions and understandable tasks to exert their own capabilities.Because of this,task allocation plays an important role in the information flow and resource cir-culation of the unmanned swarm system,and is the power source for the formation of the unmanned swarm system’s capabilities.To this end,this thesis orients to the task allo-cation problem of unmanned swarm based on the current theoretical and technical level.From the perspective of task allocation mode and method,the main work includes the following four aspects:Firstly,in view of the current task allocation model’s over-reliance on ground center operators,the task allocation model based on man-machine coordination is studied.First,aiming at the shortcomings of traditional command and control methods that are difficult to fully utilize the autonomous intelligence of unmanned platforms,the command and control method based on platform intelligence is studied to make full use of the limited Artificial Intelligence of unmanned platforms to serve task assignment.Secondly,the task-oriented human-machine collaboration model is studied,and how the intelligence of the operator and the intelligence of the unmanned platform can cooperate to complete the task assignment.Then,according to the part of the work that the platform intelligence needs to complete,the autonomous response method of unmanned cluster tasks is stud-ied.Furthermore,the task allocation process based on human-machine collaboration is analyzed.Secondly,aiming at the problem that the current pre-task allocation method can-not meet the operational requirements of unmanned swarm,a pre-task allocation method for unmanned swarm is proposed.Constraints such as heterogeneous platforms,diverse tasks and large scale in unmanned swarm make it difficult to solve the pre-task allocation problem.In order to solve this problem,the basic model of pre-task allocation problem is constructed.Then,based on the genetic algorithm,coding rules,genetic operators,decod-ing solutions,supplementary rules,and other contents are studied for the problem model to complete the design of improved genetic algorithm.Furthermore,the pseudo codes of task initialization,crossover operator and mutation operator are given respectively to ex-plain the algorithm.Experimental results show that the proposed algorithm can optimize the solution under the background of small-scale and large-scale problems and has good convergence.Thirdly,aiming at the challenges of dynamic task assignment in unmanned swarm,a multi-agent distributed autonomous assignment method for dynamic task assignment in unmanned swarm is proposed.Task assignment of unmanned swarm is faced with chal-lenges such as difficult access to global state information,limited communication ability and high dynamic change.According to the characteristics of dynamic task assignment problems,a theoretical model is constructed firstly.Secondly,a distributed autonomous negotiation mechanism for task assignment is designed based on contract network mech-anism and multi-agent mechanism.Further,a method of distributed autonomous assign-ment of tasks is proposed,which can realize the requirements of decentralization,low communication overhead and strong real-time performance.Experimental results show that the proposed methods can effectively accomplish the dynamic task assignment of unmanned swarm under distributed conditions.Fourthly,to solve the problem of platform damage caused by high threat in the bat-tlefield environment,a new unmanned swarm damage-tolerant task assignment method is proposed.In the battlefield confrontation environment,all kinds of fire interwoven and electromagnetic environments are extremely complex,which makes the unmanned platform prone to being damaged,failure and loss of contact,which seriously affects the performance of the swarm system.Based on this background,a damage-tolerant system model of swarm system and a task backup overlap mechanism are proposed to improve the success probability of tasks by using the task backup mechanism.At the same time,the task overlap mechanism is used to reduce the extra resource cost caused by backup.Then,a new method of task assignment is proposed.Experimental results show that the proposed methods can effectively improve the success probability of the task with less additional resource overhead.