Guidance for Autonomous Precision Landing on Atmosphereless Bodies

Abstract

Two key technologies that will likely fly on next-generation planetary landers are hazard detection and avoidance (HDA) and vision-based navigation. The purpose of HDA is to make previously unsafe landing sites accessible by future vehicles, and to increase the safe landing probability overall. Vision-based navigation will enable to target specific landing sites more precisely and also plays an important role for HDA. This paper formulates requirements on the guidance system that derive from the needs of these new systems. After the introduction of a reference mission scenario with HDA and vision-based navigation inthe-loop, an extensive literature survey of the state-of-the-art in approach-phase guidancealgorithms is presented. This presents the methods under the angle of HDA-compatibility. The sheer number of papers mentioned here leads to the conclusion that trade-offs are challenging. It is recommended to do a down-selection from the papers presented here, and do a performance-based trade study for the particular test case under consideration with a few candidates. As an example, E-Guidance is presented for a Mercury-and a Moon-landing scenario. Because the results differ dramatically, it is concluded that it is essential to perform trade-offs on the actual mission scenario under study, and that trades cannot merely be based on references in the literature. Because the proposed E-Guidance method is not satisfyingly robust, it is a goal for future research to improve upon this.

A lander mission to probe subglacial water on Saturn׳s moon Enceladus for life

“…12. Illustration of a top-level GN&C system architecture [83]. on minimizing a performance index that depends on the weighted flight time and commanded acceleration.…”

A semi-analytical guidance algorithm for autonomous landing

“…Traditional closed-form guidance schemes (such as Apollo guidance, E-guidance) are sub-optimal and do not include explicitly path constraints, potentially leading to infeasible trajectories. On the other hand, fully numerical methods, although extremely flexible in terms of optimality and constraints evaluation, require the handling of tens of optimization variables or complex gradient based optimization techniques, and thus they are computationally intensive (Gerth and Mooij, 2014). With the proposed approach, the feasibility region is increased with respect to traditional polynomial algorithms, avoiding at the same time the higher computational cost of complex optimization methods, being the number of variables to be optimized very low.…”