Research On Task Planning For Multi-Task Based On Multiple Welfare

As China’s population ages,the care of the elderly has become a major social concern.In order to better care for the elderly and other people with mobility problems,service robots have gradually become the choice of the public.Compared to single robots,multi-robot systems are robust,efficient and easily scalable,and have greater advantages for handling complex tasks.In this thesis,we propose a multi-robot task planning system applied to the nursing home context based on the laboratory’s intelligent room model for helping the elderly and disabled,and divide the system into two parts: task assignment and path planning.In order to better fit the actual nursing home context,so that different functional welfare robots are responsible for different types of tasks,the tasks in this context are divided into five categories according to the function of each robot and the corresponding service duration.To solve the multi-robot task assignment problem,a multi-objective optimization model is established by combining the characteristics of the nursing home context.With the minimum total time spent to complete the task,the minimum amount of energy consumption of the system and the maximum number of tasks completed as the optimization objectives,and the three optimization objectives are transformed into a single objective by linear weighting method;The constraints of practical problems such as timed return demand,insufficient energy reserve,area capacity limitation and task repetition assignment are also considered,and the task completion time constraint,energy consumption amount constraint,task load constraint and number of robots constraint are set for each robot respectively.The augmented objective function composed of the objective function and constraints is used as the fitness function to solve the task assignment scheme using an improved particle swarm algorithm.The algorithm integrates the updated formulas of fireworks algorithm,genetic algorithm and particle swarm algorithm to improve the exploration ability of the algorithm.In the practical solution,an encoding method is designed to transform the decision variables from discrete to continuous type,which enables the continuous type algorithm to solve discrete problems and transfer the processing of some constraints from the encoding layer to the decoding layer.After the multi-robot system completed the task assignment,the experimental model of the raster environment was designed according to the task sequence as well as the actual map environment.Taking the total distance traveled by all robots as the performance index,and considering the constraints of not colliding with other robots and obstacles,the A* algorithm is used to solve the path planning problem in a static environment and plan a collision-free optimal path for each robot.Since the A*algorithm cannot solve the unknown obstacle and local optimum problems in dynamic environments,a dedicated strategy for dynamic rescheduling is designed to improve the A* algorithm and mix the A* algorithm with the artificial potential field method,using the ensemble potential field function of the current location of the node as the heuristic function,and the planned paths are free of blocking phenomena and each robot effectively avoids unknown obstacles.In summary,the designed algorithm can effectively solve the multi-robot task planning problem in this nursing home context.Compared with the original algorithm,the improved particle swarm algorithm improves the accuracy of solving the task assignment problem by 1.74 times,the path length of the route and the number of expansion nodes required to solve the dynamic path planning problem by the improved A* algorithm are reduced by a maximum of 7.42% and 1.53 times,respectively,and the algorithm running time is reduced by a maximum of 0.14.This multi-robot task planning system can be applied to the nursing home context.