Graph-based cooperative navigation using three-view constraints: Method validation

Abstract

One of the hard issues that arises in distributed navigation is keeping an up-to-date and consistent estimation of the dependency between the solutions computed by each one of the involved agents. This issue is critical for the consistent information fusion in distributed cooperative navigation and was recently tackled using a graph-based approach for the on-demand calculation of cross-covariance terms. In particular, the approach was applied to a method for visual aided, distributed cooperative navigation based on three-view geometry constraints, in which a measurement is formulated whenever the same scene is observed by several robots, not necessarily at the same time. The purpose of this paper is twofold. First, the claim that on-demand calculation of cross-covariance terms in three-view-based cooperative navigation is further substantiated, and the difficulties with other existing techniques are emphasized. Second, the efficiency of using the on-demand calculations is validated by comparing the results to those obtained by assuming the three-view multi-robot measurements schedule is known a priori. In this latter method, the required cross-covariance terms are calculated using a fixed-lag centralized smoother. The comparison clearly shows the advantages of using the on-demand scheme.