Fast UAV Trajectory Optimization Using Bilevel Optimization With Analytical Gradients

In the article, we present an efficient optimization framework that solves trajectory problems by decoupling state variables from timing variables, thereby decomposing a challenging nonlinear programming (NLP) problem into two easier subproblems. With fixed, can be optimized efficiently using convex optimization, and in separate NLP, which forms bilevel problem. The challenge of obtaining gradient is solved sensitivity analysis parametric NLPs. exact analytic computed dual solution as by-product, whereas existing finite-difference techniques require additional optimization. optimizes both demonstrated on generating trajectories for UAV. Numerical experiments demonstrate converges significantly more reliably than standard NLP solver, analytical gradients outperform finite differences terms computation speed accuracy. Physical its real-time applicability reactive target tracking tasks.