A modified HP-adaptive pseudospectral method for multi-UAV formation reconfiguration

Abstract

The muti-UAV formation reconfiguration is a challenging problem due to the state constraints. This is because state constraints usually lead to a large-dimensional nonlinear program (NLP), which highly increases the computational burden. Inspired by the idea of an exact penalty function scheme, a novel computational method is developed by modifying hp-adaptive pseudospectral method (PM). More specifically, hp-adaptive PM is introduced to transform the UAVs reconfiguration control problem into an optimization problem, while the exact penalty function is utilized to remove the inter-UAV distance state constraints. The original problem is then converted into a standard NLP problem which can be handled by the gradient-based method with the off-the-shelf software packages. Comparing with the hp-adaptive PM, the proposed method dramatically decreases the computational overhead. Numerical simulations are provided for illustrating the effectiveness of the proposed scheme.