Research On The Robot Cooperation Problem Of Multi-robot System

With the development of society, the field of robotic applications is extending. However, limitations of single robot system exist in information acquisition, processing and control ability. For complex tasks and changeful environment, the ability of single robot system is more insufficient. Compared with single robot system, multi-robot system (MRS) has obvious advantages. Through multi-robot cooperation can efficiently complete complex tasks which single robot is unable or difficult to do. As one of the hotspots in the field of MRS in recent years, the problem of multi-robot cooperation belongs to the category of organizational problem and relates to the basic theory of game theory, operations research, artificial intelligence and other fields. Through the research of cooperation, it can better realize the organization and control of multi-robot groups, and can also provide a useful reference for the research of artificial life, social science and other issues. Therefore, the subject research has important scientific significance and application value.From the view of quantitative analysis, the problem of multi-robot cooperation includes generation of total task, task decomposition, task allocation and so on. On the basis of completely collecting and analysing domestic and foreign literature on this topic, we conduct a more in-depth research on multi-robot cooperation and its application in MRS. The main research work is as follows:(1) Generation and decomposition of total task in MRSAccording to the task decomposition problem of rescue robot system under dynamic searching environment, based on a brief analysis of total task generation mechanism of the system, k-mean algorithm is introduced into the task decomposition problem of multi-robot cooperative environment exploring. We propose the approach kMC-RCRSTD based on k-mean clustering which is used in task decomposition of RoboCup rescue system. The simulation results show that kMC-RCRSTD is superior to all other approaches for its good performance in task decomposition and advantages of low rate of off-group points, small subtask difference degree, low time complexity, etc. Aiming at the task decomposition problem of soccer robot system under dynamic adversarial environment, based on a brief analysis of total task generation mechanism of the system, we present the approach CDSGF-SRSTD based on court division and soccer gravitational field which is used in task decomposition of soccer robot system. The simulation results show that CDSGF-SRSTD can improve the team competition performance and, compares with similar approaches, effectively shorten task decomposition time and increase efficiency of task allocation.(2) Multi-robot task allocation (MRTA) model based on robotic time utility functionAccording to the shortage that existing approaches just use algorithm optimization and do not quantify the task allocation results, on the basis of taking into account these factors such as distance between robot and task, angle difference from initial to target pose of robot, and robot’s actions after reaching the target point, a MRTA model RTUF-MRTA based on robotic time utility function is addressed. The model defines and constructs robotic time utility function, makes theoretical analysis for the solvability of robotic time utility function, gives the existence and uniqueness theorem of the best task allocation scheme and proves it. Relevant examples verify the validity of the model. The model proposed provides solid theory safeguard for the future research of this dissertation.(3) MRTA based on genetic algorithmAiming at task allocation the NP-hard problem, on the basis of giving the time utility value matrix of robots relative to tasks, we propose the approach GA-MRTA based on genetic algorithm for MRTA. Experimental results of RoboCup rescue simulation task allocation show that the problem solving and space optimization based on GA-MRTA can realize the best mapping from robot to task and enable the system to achieve more satisfactory task allocation effect.(4) MRTA based on modified particle swarm optimizationAccording to the disadvantage of slow convergence speed and weak global optimization ability of genetic algorithm, in further, the MRTA approach MPSO-MRTA based on modified particle swarm optimization is presented. Simulation results of RoboCup 2D soccer robot task allocation show that the problem solving based on MPSO-MRTA not only can realize the best mapping from robot to task, but also has some advantages such as high algorithm convergence rate, fast convergence speed, strong global optimization ability, etc.(5) MRTA based on multi-objective optimizationOn the basis of considering time and energy which are usually important index for evaluating task complete effect in MRS, we first propose the MRTA approach MOO-MRTA based on multi-objective (time utility, energy utility) optimization. The approach gives the definition and construction of robotic energy utility function, establishes task allocation model based on multi-objective, discusses the model solution and other issues. RoboCup rescue simulation experiments show that the approach has less computing time, can quickly find the set of Pareto-optimal task allocation scheme and enable rescue team to get the desired competition results.In the final part, we make a summary of the dissertation, describe the primary innovations and achievements of the research, and point out the problems to be further studied.