Study On Architecture And Data Consistency Algorithms In Autonomous Mobile Robot

Unmanned ground vehicle (UGV) remains a hot spot in intelligent robots research lit-erature. The architecture of UGV combines modules of perception, fusion, modeling, plan-ning, reasoning and behavior. It makes autonomous navigation in complex environments possible. During these years, distributed system has become an attractive research domain due to its fabulous expandability and high fault tolerance. In the framework of distributed system, all of agents are modeled as a distributed computing system. Therefore, How to make the data scattering on distributed nodes consistent becomes a significant problem so as to guarantee the whole system is working properly. Due to reasons mentioned above, we concentrate on the distributed system architecture and the data consensus problem.To begin with, this thesis studies the history of the development of UGV, discusses multiple classic architectures, introduce the consensus algorithm in the context of dis-tributed systems. Based on that, this thesis proposed a distributed UGV architecture, which models function modules as agents, and classified them into three groups:perception mod-ule, processing module and behavior module. All of agents are built on a connected envi-ronment, which provides the following functions:data/message transferring, global states broadcasting and consensus maintaining. These agents were distributed on different com-puting nodes, each is working independently. According to the requirements from goal and environments, agents are organized as an asynchronous navigation pipeline to perform nav-igation task effectively. This architecture can be naturally extended to multi-agents system, and provide the reciprocal operations among systems. Empirical results in wide range of environments verify that the proposed architecture is robust and extendable.Maintaining the data consensus in multi-agents connection environment is of great importance in robots system architecture. This thesis analyzed three classic data consen-sus algorithms, and proposed a new data consensus algorithm that applying to distributed robots system. This algorithm divides the consensus requirements to three layers:First, the consensus in a single robot, it guarantees the consensus between the main agents and the backup. When main agents crash, the backup can replace the failed agents seamlessly to prevent the halt of the whole system. Second, the consensus among multiple robots, it requires maintaining data consensus while sharing and updating data among multiple robots.Tests on simulation and real-system show the hierarchical consensus algorithm is adept to task performing in multiple robots system.Finally, this thesis makes the conclusions and suggests directions for future research.