RECEN:Resilient MANET Based Centralized Multi Robot System Using Mobile Agent System

摘要

In order to improve the efficiency of a Multi-Robot System (MRS), various algorithms are proposed including centralized and decentralized algorithms. Centralized algorithms can maximize coordination efficiency as it has global information. However, these algorithms have few critical problems such as a single point of failure (SPOF) problem. To the best of our knowledge, no papers directly tackle this issue but propose decentralized approaches, which can only achieve suboptimal efficiency. In this paper, we provide RECEN, a new resilient Mobile Adhoc NETwork (MANET) based centralized architecture using a Mobile Agent system (MAS). A mobile agent is a software process that can migrate from one machine to another while continuing its tasks on the new platform. Applying this technique in MRSs, a master agent can migrate to different robots when facing hardware issues while continuously manages its slave robots efficiently with Multi-Robot Task Allocation (MRTA) algorithm. In order to verify improved resilience by RECEN, our simulation compares RECEN with a general centralized MRS. The result shows that the performance of RECEN outweighs a general centralized MRS by completing 10000 (maximum number of tasks) and 141.42 tasks respectively with 10 robots when there is no sudden accident that causes robot failure before migration. Taking such accidents into account, the simulation result reports that RECEN completes 2768.34 tasks on average with the same number of robots. Moreover, our multi-robot surveillance application in real world environment demonstrates resilience of RECEN by continuously managing slave robots despite hardware issues in a master robot.