Agile Autonomous Guidance using Spatial Cost-To-Go Functions

This paper describes a guidance scheme for interactive spatial guidance tasks involving complex environments and agile vehicles. Receding Horizon (RH) control is an established control methodology which has been used successfully for many control applications. More recently it has been employed for autonomous vehicle guidance. Its success hinges on using an appropriate cost-to-go (CTG) function that accounts for the discarded tail of the trajectory. In this paper we describe a RH trajectory optimization scheme based on a spatial, statedependent CTG function. The spatial CTG is computed as the approximate value function associated with the minimum time optimal trajectory planning problem. The approximation is based on a finite-state model of the vehicle dynamics that captures the feasible flight envelope using a small set of motion primitives. This simplified model allows efficient computation of the CTG function over large geographical spaces while taking into account the critical interaction between the vehicle dynamics and environment. Besides enabling computational efficiency this scheme provides a more consistent approximate of the infinite horizon using RH and also facilitates taking into account complex environment and future extension for onboard sensing. The paper describes in details the computation of the spatial CTG function, the guidance system integration, and the experimental implementation and demonstration in our Interactive Guidance and Control Laboratory.