Abstract:

In recent years， the intelligent unmanned cluster system has attracted extensive attention. The cooperative navigation of multi-agent nodes in the cluster is a key problem to realize complex cooperative control. To overcome the possible problems of asynchronous and discontinuous measurement in the complex occlusion environment， a distributed collaborative navigation method is proposed based on graph optimization algorithm. The factor graph model is established based on the state constraints composed of the relative ranging information between nodes provided by （Ultra Wide Band， UWB） sensors and the dead reckoning information of nodes based on inertial sensors. The global optimization characteristics of graph optimization algorithm are used to solve the problem that the traditional filtering methods cannot adapt to non synchronization of measurement. Through the adaptive dynamic distributed topology， the measurement between nodes can be fully used to solve the problem of insufficient measurement of anchor points in the complex environment. Based on the real-time estimation of gyro bias， the individual vehicle’s ability to maintain measurement accuracy without external measurement in a short time is improved. The mathematical simulation and experimental results show that in the complex occlusion environment， the proposed method has the ability to process the measurement of dynamic addition and deletion and time asynchrony， and can estimate and compensate the gyro bias in real time. It can improve the positioning accuracy retention ability of a single node in the short-term communication rejection environment by 61%， and the comprehensive positioning accuracy of the cluster in the complex environment of limited communication is doubled.

Key words: collaborative navigation, unmanned cluster, graph optimization, factor graph, multi-agent