Overview

The Planning and Scheduling Group builds automated planning and scheduling systems for NASA missions. These planning and scheduling systems are essential components of autonomous spacecraft, deep space probes, planetary rovers, and autonomous vehicles. They are also integral elements of ground operations tools for space missions. Planning and Scheduling Group activities include: Conducting basic research into computational methods used in automated planning and scheduling systems. Designing and implementing automated planners as components of autonomous systems. Designing and implementing planning and scheduling tools used by mission operators.

Project List (Active)

Activity Planning and Sequencing for Mars Science
Project Lead: James Kurien

Constraint Based Planning
Project Lead: Conor McGann
Development of planning techniques and software for domains with complex temporal and resource constraints.

Decision Theoretic Planning for Planetary Exploration
Project Lead: Nicolas Meuleau
This projects aims at producing high-quality contingency planning and re-planning solutions by scaling-up decision theoretic techniques to real NASA problems. We use planetary rovers as a test-bed and focus onthe three following issues: structured and concurrent planning domains, continuous uncertain state variables, oversubscription.

Europa OSTPV
Project Lead: Jeremy Frank

Game Theoretic Scheduling of the Deep Space Network
Project Lead: Jeremy Frank
NASA missions over the next two decades will require communications bandwidth that exceeds current Deep Space Network capacity by an order of magnitude or more. An important open issue in the design of the DSN array is how such large arrays can be efficiently scheduled. Both the new and current DSNs are key resources for NASA missions, and the use of the DSN is the subject of intense political negotiations. In these negotiations each mission has a clear objective (getting their science data down) and the DSN operators have an objective (not overloading the system and serving their customers). However, it is overly simplistic to view the emergent process as a search for a single global objective. A more accurate and useful description is to take a game theoretic approach and view the process as a search for either a Pareto optimum or a Nash Equilibrium One major liability of the current DSN scheduling system is that this "game" is competitive and playing it requires significant staffing on the part of each mission and the DSN operations team. In this project we argue that by taking a game theoretic view, and utilizing and extending scheduling technology, we can put in place a system for negotiating DSN schedules that will both handle the additional complexity of scheduling DSN arrays and simplify the scheduling game - thus reducing the staffing requirements for DSN scheduling.

Mixed-Initiative Planning
Project Lead: John L Bresina
Our research will explore fundamental issues of mixed-initiative planning and scheduling that will address these shortcomings by focusing on the following two challenges:Preferences: Enabling the user to specify preferred solution characteristics and planning advice in order to influence the plan search and resultant solution. Explanations: Providing effective summarizations and comparisons of solutions, to enable trade-off analyses, and providing explanations of planning decisions and failures. The different types of preference will be unified and operationalized within an optimizing search approach (e.g., branch and bound), which will support trade-off analyses.

SOFIA Observation Scheduling
Project Lead: Jeremy Frank
Development of observation scheduling and flight planning techniques for the Stratospheric Observatory for Infrared Astronomy (SOFIA) airborne observatory.

Spacecraft Autonomy for Vehicles and Habitats
Project Lead: Ari Jonsson

Universal Executive (PLEXIL)
Project Lead: Ari Jonsson
We are developing a language, called PLEXIL, that is designed specifically for flexible and reliable command execution. It is designed to be portable, lightweight, predictable, and verifiable, and at the same time it is very expressive. Project leads: Ari Jonsson and Vandi Verma

Project List (Inactive)

Remote Agent
The Remote Agent Experiment was the first instance of state of the art artificial intelligence system being given primary command of a spacecraft. The Remote Agent software operated NASA's Deep Space 1 spacecraft and its futuristic ion engine during two experiments that started on Monday, May 17, 1999. For two days Remote Agent ran on the on-board computer of Deep Space 1, more than 60,000,000 miles (96,500,000 kilometers) from Earth.
+ Visit Remote Agent