Vandermonde Trajectory Bounds for Linear Companion Systems (2302.10995v1)

Abstract: Fast and accurate safety assessment and collision checking are essential for motion planning and control of highly dynamic autonomous robotic systems. Informative, intuitive, and explicit motion trajectory bounds enable explainable and time-critical safety verification of autonomous robot motion. In this paper, we consider feedback linearization of nonlinear systems in the form of proportional-and-higher-order-derivative (PhD) control corresponding to companion dynamics. We introduce a novel analytic convex trajectory bound, called $\textit{Vandermonde simplex}$, for high-order companion systems, that is given by the convex hull of a finite weighted combination of system position, velocity, and other relevant higher-order state variables. Our construction of Vandermonde simplexes is built based on expressing the solution trajectory of companion dynamics in a newly introduced family of $\textit{Vandermonde basis functions}$ that offer new insights for understanding companion system motion compared to the classical exponential basis functions. In numerical simulations, we demonstrate that Vandermonde simplexes offer significantly more accurate motion prediction (e.g., at least an order of magnitude improvement in estimated motion volume) for describing the motion trajectory of companion systems compared to the standard invariant Lyapunov ellipsoids as well as exponential simplexes built based on exponential basis functions.