Active cooperative observation of A 3D moving target using two dynamical monocular vision sensors

Vision-based 3D moving target observation (3D-VMTO) is a central problem in the field of mobile robotics. In this paper, a new 3D-VMTO method called active cooperative observation is proposed through cooperation among two dynamical monocular vision sensors (MVSs). Under this method, an algorithm based on an extended set-membership filter is designed to fuse observations from different MVSs. An optimal observation condition is then introduced as the constraint into the relative velocity coordinates based path planning scheme, so as to reduce the influence of relative positioning among MVSs and their target on cooperative observation results. Experiments on a multiple rotor-flying-robots testbed demonstrate the validity and feasibility of the proposed method