Adaptive event-triggered actor–critic algorithm for optimal 3D formation circumnavigation with relative measurement and an unknown moving target

Abstract

This paper studies the optimal three-dimensional (3D) formation circumnavigation problem for multiple unmanned aerial vehicles (UAVs) with relative measurement and an unknown moving target. First, the 3D formation circumnavigation errors under the relative measurement and directed topology, that is, the spherical enclosing along the normal direction, the latitude surrounding along the north direction, and the longitude formation along the east direction, are constructed. Then, radial basis function neural networks (RBFNNs) based adaptive estimation is designed for each UAV to approximate the unknown target velocity using relative information. Further, to achieve the optimal energy of 3D formation circumnavigation along the normal, north, and east directions with lowered data transmission cost, an event-triggered actor–critic algorithm is designed under limited neighbor and relative target information. The stability of the whole system is analyzed under continuous sampled and event sampled cases, respectively, and the Zeno phenomenon is excluded. The effectiveness is validated through simulation under cases of with and without random noise. Meanwhile, comparative simulations are performed to validate the system energy and convergence time of the proposed method in contrast with two relevant works.