Optimization Methods for Automated Space Mission Planning

Thomas Fletcher Cunningham, Purdue University

Abstract

Activity planning for space mission operations has traditionally been a human-in-the-loop effort, conducted by ground operators. Over the past two decades, advances have been made toward automating the mission planning process, in an effort to improve the efficiency of the mission operations system, while increasing the mission return. In keeping with NASA’s goals, some aspects of onboard mission planning are increasingly used for complex missions, particularly for planetary surface missions that are subject to long communication delays. This dissertation research develops an automated mission planning framework and applies it to two spacecraft scenario case studies: a science orbiter and a science rover mission. Mission plans are optimized on the basis of science return, accommodating spacecraft movement to sites of scientific interest according to ground-team preferences, while staying within rover engineering and traverse-related constraints. Automated mission planners offer the capability to schedule engineering and science activities onboard, without ground-in-the-loop interaction. Resource modeling and path planning can be done onboard, reducing the need for modeling and verification by ground operators. Further, automated mission planners may incorporate an optimization executive that maximizes the mission return within the available resource constraints. The proposed planners may be utilized onboard autonomous spacecraft and rovers with limited human support. Also, they may be run on the ground by mission planning teams to provide additional insight during the planning process. Utilizing a variety of optimization approaches, the developed automated mission planners establish the planned sequence of activities, including and engineering activities, while adhering to constraints imposed by orbital geometry or planetary pathing requirements and resource availability. The focus of the work is on remote, robotic missions in which human-in-the loop decision input is delayed or at times unavailable. Two major classes of robotic missions are examined: Orbital science missions in which primary science activities are performed periodically at a specified rate, and a planetary rover mission in which a larger variety of science activities are interspersed with unique terrain navigation activities. The automated mission planning framework is designed to be adapted based upon the application. Optimization methods suitable for different mission planning problems are presented, comparing methods on the basis of computation speed, resources required and solution value. The Aerospace Systems Engineering definitions for “robustness” and “flexibility” are given quantifiable, mathematical definitions and are incorporated into the framework as quality metrics to provide criteria with which to evaluate and compare the produced activity plans. The metrics “reliability” and “latent performance index” provide additional criteria for plan evaluation. A variety of automated mission planning algorithmic approaches are developed and described functionally and mathematically. Planning tools capable of plan verification, Monte Carlo simulation-based verification and plan variation analysis are developed and described in detail. Two detailed, step by step case studies are developed, applying and running all the mission planning and analysis tools to provide planning solutions and analysis of generated plans for the science orbiter and science rover scenarios. The application of the developed planning solutions to the presented missions, including the determination of the quality metrics, are seen as the primary contributions to the advancement of the state of the art in automated mission planning. The Automated Mission Planning and plan analysis techniques and practices are summarized into a User’s Guide to Automated Mission Planning.

Degree

Ph.D.

Advisors

Marais, Purdue University.

Subject Area

Artificial intelligence

COinS