Fast, Robust Real-Time Trajectory Generation for Autonomous and Semi-Autonomous Nonlinear Flight Systems

"Our approach is to decompose the trajectory generation problem intothree distinct, but tightly coupled pieces: waypoint path planning (WPP),dynamic trajectory smoothing (DTS), and adaptive trajectory tracking (ATT).The WPP plans paths at a high level without regard for the dynamicconstraints of the vehicle. This affords a significant reduction inthe search space, enabling the generation of extremely complicatedpaths that account for pop-up threats and dynamically changingthreats.The essential idea of the DTS is to give the trajectory generator asimilar mathematical structure as the physical vehicle. The DTS usesa simple, but novel algorithm to generate smoothed trajectories inreal-time without performing any on-line optimization. Thetrajectories that are generated by the DTS have the same path length as thewaypoint path generated by the WPP and also minimize the deviationfrom the waypoint path.The third step of our approach uses adaptive backstepping to transformthe trajectory generated by the DTS to a feasible trajectory that canbe followed by an autopilot with appropriate velocity, altitude andheading commands. The proposed approach is computationally efficient: it can handle hundreds of threats, including pop-up threats. It does not require on-line optimization. Is very well suited to applications with timing constraints. Planning can take place at the waypoint level, where it is trivial to calculate path length, and therefore estimated time-of-arrival (ETA). The trajectories can be represented in a compact fashion, in both space and time. In particular, this will allow higher-level task planning algorithms to reason about the feasibility, or desirability of different trajectories."