A Study On Multi-Robot Cooperation

With a large range of advantages, such as robustness, adaptivity, distribution on time and space, high efficiency, a multi-robot system(MRS) can outperform a single robot and tends to be accepted in many applications where a single robot system has been thought not suitable. MRS can be widely used in many fields, including astronautics, national defense, industry, social service and so on. By cooperation between robots, a group of robots can be more capable than simply gathering all robots together, of solving all kinds of complex tasks, so how to design the cooperative framework in robots group is the key problem in the multi-robot system study. In this dissertation, We focus our research mainly on the problem of task allocation, which is part of the cooperation issue in MRS.Multi-robot task allocation(MRTA) has recently risen to prominence, and become a key research issue in multi-robot domain. Over the past decade, a large number of work about multi-robot cooperation has been made, but most of work has addressed only some specific task or application, and cooperation architectures having been proposed to solve task allocation problem has been short of generality. This dissertation investigates the problem of MRTA by building general-purposed cooperation framework at the level of task allocation, using intentional cooperation and swarm cooperation(emergent cooperation) respectively. With intentional cooperation, we think that the description of task and robot should be formulated by abstracting robots or tasks' fundamental properties and constituting them into a model, and cooperation approached must satisfy the requests of logic, generality, and practicability. With swarm cooperation, we believe that a intrinsic and proper organizing mechanism is very important for a group of robots, especially in large scale. Looking around the world where we live, we found that in many social groups, balancing in profit, quantity or other ways is a common principle or law, and keeping this principle makes, the group will live stably, or will corrupt. With this finding, a new multi-robot organizing principle is put forward, and we call it balancing principle. Focusing on above thoughts, the main contributions of this dissertation are:1. The introduction of research contents and current status of MRS is given. Many research achievements on intentional cooperation and swarm cooperation are analyzed and compared. A hierarchy of cooperative goal of MRS is proposed.2. The constraints of tasks on cooperation of MRS is analyzed, and the effect of learning fully the characteristics of tasks on control structure, cooperative mechanism, and communication mode in MRS are emphasized. Two kinds of task, loosely-coupled tasks and tightly-coupled tasks are described. Based on an analysis of MRTA problem, a classification on MRTA is given. The progress in MRTA and two approaches are discussed.3. Three design rules for task allocation architecture are defined. Four strategies for task selection are suggested, and with those, a new task allocation framework is put forward, which is based on a mathematically-modeled fitness, and analysis on the framework is also made. The validity, robustness and flexibility are demonstrated by applying this task allocation approach into the multi-robot mission of waste cleanup.4. Based on the analysis of MRS with self-organization and it's difficulty in description, macroscopic describing is designed as a new method to describe the self-organizing systems, and then some characteristics of the describing way are analyzed; By Using this description, a dynamical model has been established for a self-organizing MRS facing a tightly-coupled mission. With the model, a self-organizing principle is built correspondingly, and a MRS with task allocation mechanism using the principle is developed, and simulation is carried out by applying the system into a mission of object transportation.5. Balancing concept and principle are proposed, and the mathematical definition of balancing degree metric is given. In order to use balancing principle, a behavioral model for robots is built. The relationship between balancing and self-organizing phenomena in Multi-robot system is analyzed, which proves the significance of balancing principle in MRS self-organization. Based on balancing principle, an algorithm of task allocation for multi-robot system is put forward.6. The balancing principle based MRTA algorithm is studied through using it in the mission of object moving by a group of mobile robots, and in the study, deterministic behavioral model and stochastic behavioral model are adopted for robots respectively. The double token technology is introduced into the algorithm as a mechanism for avoiding behavioral conflicts between robots.7. Addressing the task of cooperative multi-robot observation of multiple moving targets, the application of balancing principle into personality evolution in MRS is investigated, and two new tracking methods PE-CMOMMT and PEL-CMOMMT. Simulation results in the validity of the two methods.By simulation, it is showed that the multi-robot cooperative architectures put forward in this dissertation have high coherence, robustness, adaptivity and generality. Balancing principle provides a new way to solve the control problem of multi-robot system from the view of system' intrinsic structure.