Research On Dynamic Task Allocation In Cooperative Multi-robot System

In recent years,multi-robot systems have increasingly been applied to military,unknown environments and disaster areas.In addition,such fields as production automation,intelligent warehouses and take-away catering are all areas in which robots can put expertise to good use.Multi-robot systems improve the performance of individual robot by collaborating and sharing information,such as task execution efficiency,robustness,flexibility and fault tolerance,including distributed decision-making,formation control,area coverage and related applications.As the most central module in the design and implementation of multi-robot system,the effect of task assignment directly affects the system performance and work efficiency.Especially in recent years,the problem of dynamic task assignment for multi-robot has drawn more and more attention and become the most important hot topic in multi-robot research.In the multi-robot system application,task assignment is divided into static task assignment and dynamic task assignment according to the task situation.Static task assignment means that the task information is known to the robot and remains stable before the task is executed;For dynamic task assignment,robot task assignment is a dynamic process,and needs to be constantly adjusted and changed according to the change of task environment and the demand of subgroups for interests.However,most of the multi-robot studies are aimed at static task allocation,and the distributed dynamic task in unknown environment is only a preliminary model,and the allocation strategy is not perfect.Moreover,most researches of multi-robot system do not proceed from the perspective of profit when optimizing the allocation results,and rarely touch upon the multi-robot global optimization problem.The problem of task assignment under the dynamic environment of cooperative multi-robot system plays an important role in the research of multi-robot collaborative control.In this paper,a multi-robot cooperative fire-fighting model is established for distributed dynamic task assignment.In this model,factors like time,energy and fire are considered and been quantified.After all the factors are combined,a distribution strategy based on market mechanism and in line with the task assignment of dynamic fire fighting is proposed.Then the model is modeled in the simulation platform according to the corresponding model,and simulation results show that this task allocation strategy can efficiently optimize the assignment of tasks in different situations.Aiming at the optimization problems existing in the dynamic task assignment of multi-robot systems,this paper proposes an improved task assignment algorithm using Pareto improvement based on the initial assignment of contract net.When multi-robot systems perform fire fighting tasks in parallel,the multi-robot is firstly divided into several sub-groups by initializing task assignment.Then each subgroup contracted a firefighting task,and the subgroups performed Pareto improvement with the nearby subgroups at the same time to determine the robots to be migrated to achieve the Pareto optimality between two subgroups.Finally,Pareto improvement is applied to all subgroups to achieve global Pareto optimality by using the traversal of binary tree.Theoretical analysis and simulation results show that compared with the traditional contract net method,Pareto improved algorithm can accomplish the task of fire rescue efficiently in time,which is obviously superior to the traditional contract net method in terms of system benefits.And the time spent on task has been significantly reduced,compared with learning algorithm and ant colony algorithm.