A Multi-channel Allocation Mechanism Design For Multi-robots Coordination

The Multi-robot coordination is a hot research topic in the field of distributedartificial intelligence, one of the key challenges is to establish an effective mechanism toachieve rapid information sharing between the robots, via the communication channel.In the decentralized multi-robot system, each robot is only based on the current task andinstant listening channel conditions, to choose the part of the channel forcommunication and interaction, which makes the channel selected scheme may bedifferent with other robots in the wireless network. Further, constrained by the limitedcommunication coverage distance and the channel resources, the state of the system andthe state of the network, especially teammate state are not all observed to robots, onlypartly can be observed. Therefore, to establish an effective mechanism to achieve therobots’ multi-channel selection method is necessary, the new mechanism will achievefast and accurate information dissemination in the wireless network, it will effectivelyimprove the performance of the decentralized multi-robot system in coordination andcooperation.In this thesis, the author proposed a multi-channel selection method based on thesynergistic of multi-agent coordination decision. The core of the method is, the robotestablished the local observation-decision model based on limited observations of thecurrent state of the accessed channel, the neighbors, the target and the teammates. In thedevelopment of the decision-making selection strategies, the robot can only get a part ofthe state and information about the communication network and teammate robots.Therefore, we can abstract all constraints and targets as the impact elements in therobot’s local POMDP decision model, which affected the multi-channel selectionproblem, and calculated the real-time status and information obtained by the mobilerobot, it can update on the observation space locally or globally. Robot iterated the localobservation and states to optimize channel selection and meet the dynamic changes inthe network environment, and the demand of real-time transmission.The design of utility function can calculate the reward of different combinations ofthe robots, in order to achieve the global optimum. The effective selection strategies make the complex multi-robot dynamic information problem transfer into themulti-communication channel selection-decision problem; it effectively solved thebottleneck of channel resources in the decentralized multi-robot coordination, whileachieving a more targeted information distribution. This program still combinedqueuing theory, and given a simple mathematical proof of the multi-channel multi-robotselection problem. Finally, based on the theoretical design simulation, construed and testthe channel decision-selection method proposed in this paper, and analyzed of theimpact on the network in terms of network throughput, delay and network resourceoverhead.