Vision-based estimation of ground moving target by multiple unmanned aerial vehicles

This paper studies the vision-based estimation problem by using multiple unmanned aerial vehicles equipped with monocular cameras. An observer-based nonlinear filter is proposed to estimate the position and velocity of a ground moving target in real time. The observability of this bearing-only estimation problem is ensured by fusing vision measurements obtained by multiple onboard cameras. In addition, the ground moving target is modeled as a random-walking vehicle with its acceleration as an unknown, time-varying but bounded variable, and the acceleration is estimated by a projection operator. The filter design task is cast into a linear matrix inequality problem, and the solution guarantees estimation stability and allows some tolerances on nonlinear uncertainties and external disturbances. Simulation results are presented to demonstrate the performance of the proposed filter.