Constrained trajectory optimization on matrix Lie groups via Lie-algebraic differential dynamic programming (Q6930722)

This paper introduces a geometrically grounded control framework based on an augmented Lagrangian formulation of constrained Differential Dynamic Programming within the setting of matrix Lie groups. A key strength lies in the seamless integration of automatic differentiation and discrete-time Lie-group dynamics, ensuring manifold consistency without reliance on specialized derivatives. The proposed constraint-handling strategy, formulated in the error-state space, enhances robustness and addresses feasibility and active-set issues that often limit existing methods. Methodologically, the approach demonstrates strong generality, extending beyond rigid-body systems to underactuated and deformable dynamics. The experimental validation is comprehensive, covering scenarios from rigid body dynamics to quadrotor control and planar deformation, and includes performance comparisons with established solvers. The manuscript is well-structured, guiding the reader from theoretical preliminaries to practical implementation and evaluation. Overall, the work represents a meaningful advancement in constrained trajectory optimization on nonlinear manifolds, with clear implications for broader applications in geometric control.