An Inertial Dual-State State Estimator for Precision Planetary Landing with Hazard Detection and Avoidance

Abstract

The navigation filter architecture successfully deployed on the Morpheus flight vehicle is presented. The filter was developed as a key element of the NASA Autonomous Landing and Hazard Avoidance Technology (ALHAT) project and over the course of 15 free fights was integrated into the Morpheus vehicle, operations, and flight control loop. Flight testing completed by demonstrating autonomous hazard detection and avoidance, integration of an altimeter, surface relative velocity (velocimeter) and hazard relative navigation (HRN) measurements into the onboard dual-state inertial estimator Kalman flter software, and landing within 2 meters of the vertical testbed GPS-based navigation solution at the safe landing site target. Morpheus followed a trajectory that included an ascent phase followed by a partial descent-to-landing, although the proposed filter architecture is applicable to more general planetary precision entry, descent, and landings. The main new contribution is the incorporation of a sophisticated hazard relative navigation sensor-originally intended to locate safe landing sites-into the navigation system and employed as a navigation sensor. The formulation of a dual-state inertial extended Kalman filter was designed to address the precision planetary landing problem when viewed as a rendezvous problem with an intended landing site. For the required precision navigation system that is capable of navigating along a descent-to-landing trajectory to a precise landing, the impact of attitude errors on the translational state estimation are included in a fully integrated navigation structure in which translation state estimation is combined with attitude state estimation. The map tie errors are estimated as part of the process, thereby creating a dual-state filter implementation. Also, the filter is implemented using inertial states rather than states relative to the target. External measurements include altimeter, velocimeter, star camera, terrain relative navigation sensor, and a hazard relative navigation sensor providing information regarding hazards on a map generated on-the-fly.

Linear Covariance Analysis for a Lunar Lander

“…During the coast phase of flight, a star tracker will be used for rotational state estimation. Two separate sensors will be activated based on a minimum altitude threshold, a Terrain Relative Navigation (TRN) sensor 10,11 and a radar altimeter. The TRN sensor will be activated near the end of the coast phase whereas the radar altimeter will be activated during the terminal descent phase.…”

“…The radar altimeter model 11 corrupts the true altitude by measurement noise and an altimeter bias. While radar altimeters convert frequency shifts in radar signals into range estimates, the altimeter model implemented computes the altitude based on the position of the lander, which provides effectively the same information while simplifying the calculation.…”

“…The TRN sensor 10,11 is an image processing sensor that utilizes an onboard map of the landing site and overlays images captured during descent to determine the position of the lander relative to the surface. The actual implementation of an actual TRN sensor differs from the model implemented during the LinCov analysis.…”