Next: Recent Improvements to HST Parallel Scheduling
Up: Observatory Planning and Scheduling
Previous: TransVERSE: An Architecture for Configuring Astronomical Observations
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint
Kleiner, S. C. 1999, in ASP Conf. Ser., Vol. 172, Astronomical Data Analysis Software and Systems VIII, eds. D. M. Mehringer, R. L. Plante, & D. A. Roberts (San Francisco: ASP), 77
Small, Fast and Reusable: A Satellite Planning and
Scheduling System
Steven Kleiner
Smithsonian Astrophysical Observatory, Cambridge MA 02176,
Email: skleiner@cfa.harvard.edu
Abstract:
The SWAS planning and scheduling
system
provides all functions from ephemeris processing to guide star
selection to timeline generation. It embodies a dynamic scheduling algorithm
which allows schedules to be rebuilt or slipped while conserving the scheduling
effort already invested. A novel graphical interface provides interactive
scheduling. The
system uses a component design for efficient reuse.
The
software is available
for use by other missions.
The Submillimeter Wave Astronomy Satellite is 600 lb NASA Small Explorer
satellite launched into a 650 km orbit on 12/5/98. It is a radio telescope
which observes the O2, CI, H2O, 13CO and
H218O spectral lines in the 500 GHz (0.6 mm) regime from molecular
clouds in our Galaxy.
These observations will help us to understand the chemistry, energetics and
kinematics of star-forming regions.
The spacecraft was built by the NASA Goddard Space Flight Center.
The
mission is under the scientific direction of the Smithsonian Astrophysical
Observatory. The PI is Gary Melnick. The NASA Astrophysical Data Processing
program
supports the distribution of this package for other missions.
At the time of this writing, SWAS has completed its launch and
early orbit testing and is taking beautiful data. The planning and scheduling
system has worked faultlessly.
Noteworthy features of the SWAS planning and scheduling system include:
- full range of capabilities, speed and relative simplicity;
- efficient planning algorithms;
- dynamic scheduling algorithm;
- novel graphical user interface;
- a component design for convenient reuse;
The planning and scheduling system provides all the functions required to plan,
schedule and command SWAS. The system starts with a predictive ephemeris
describing the orbit and produces a timeline which is then converted to a
binary load and uplinked to the spacecraft. Between these two points the system
must compute Keplerian orbital elements, planetary positions,
Sun, Moon and Earth limb pointing constraints, target rise and set times, gyro
calibration and waypoint targets, line-of-sight velocities for LO commanding,
guide stars for the Ball CT601 star tracker, slew paths between targets,
mapping offsets and attitude quaternions. The system must then generate
commanding for the science instrument and the spacecraft data handling
and attitude control subsystems.
The system is very fast. We can generate a week's worth
of observation timelines, from ephemeris ingestion to guide star selection
and command generation, in about three minutes on a SUN Ultra-10.
The code is in C, uses Motif for the widgets, and consists of
18.5K lines of code.
We have adopted a geometrical approach to planning calculations. We
convert the Cartesian state vectors in the ephemeris file to Keplerian
elements which describe the geometry of the orbit. We calculate such things
as target rise and set times, shadow entry times, nominal roll positions,
Sun avoidance satisfaction, pointing offsets, guide star selection and
gyro calibration and waypoint target generation by clever use of
coordinate transformations. We have entirely avoided the need to step
through an orbit. We consistently use an R313 rotation corresponding
to the classical Euler angles to do our transformations.
There is a paradox in scheduling. The better the schedule,
the more fragile it generally is. A highly efficient schedule, for example,
will rarely allow one to swap targets in and out or to slip the schedule for
a week without degrading it. But the fact of real life is that
schedules must be rebuilt or replanned all the time, and usually on very short
notice. Too often a carefully crafted schedule must be thrown out in favor
of a jury-rigged replan schedule. One goal of the SWAS planning and scheduling
design was to build dynamic and conservative schedules - schedules which could
be modified or slipped without degrading their efficiency while still retaining
the scheduling effort already invested in them.
We have been able to realize this goal by redirecting our effort from building
good schedules to building good order forms. An order form is a specification
of the activities to be placed on a schedule, without
reference to the time at which the activities are to be scheduled.
The order form is a simple abstraction of the schedule, and a schedule is
a concrete instance of the order form. This is an almost trivially simple
differentiation, but it has powerful practical consequences. First, given
a good order form, it is very simple to build a good schedule from that order
form. Secondly, order forms are quite tolerant of changes. The schedules
are not tolerant of change, but given that it is easy to generate a new schedule
from an order form, we simply throw away the old broken schedules and build new
good schedules when we change the order form. It works! The rule enforced in
the SWAS planning and scheduling system is that the planner never interacts
with the schedule directly. All changes are made to the order form and a new
schedule is immediately run off from the order form.
Some SWAS planning displays are shown in Fig. 1 The all-sky
Figure 1:
SWAS planning displays such as an all-sky map,
the star tracker field of view with guide stars and a target browser which
can select targets based on different criteria.
 |
display shows how the SWAS target catalog is largely confined to the
galactic plane. The field of view display to the right shows the sky
as seen by the Ball CT601 star tracker and shows the guide stars for a given
target. The target browser allows ones to examine targets based on their
location in the sky, their visibility, the quality of their guide stars
or other criteria.
Figure 2 shows the SWAS scheduling displays. The schedule
Figure 2:
The SWAS scheduling displays showing an eight day schedule, the mixer
which controls the composition of the schedule and a visibility display showing
target rise and set times.
 |
to the left shows eight days of science observations. The waypoints which
provide safe telescope pointings between science observations are not shown.
The mixer to the right allows the planner to dynamically adjust the
composition of the order form. The schedule display is then updated immediately
to show the resulting new schedule.
Clicking on any gap in the schedule display highlights those activities
on the order
form which could be used to fill the gap. It is then simple to adjust
the order form to fill the gap. Conversely, clicking on any scheduled activity
highlights that activity on the order form, allowing one to eliminate
the activity from the schedule. One can build or modify schedules in a very
fluid and intuitive manner.
The SWAS planning and scheduling system is a set of components which can
be assembled in different configurations to do different tasks. The design
follows that of the CAMAC crates in nuclear physics which provide a
chassis, a power supply and a bus. Different modular components can then
be plugged into the crate to provide any needed functionality. We have one
single, small, well-defined global PSTask data structure which acts as a bus and
allows different
components to communicate with each other. The data structure also carries
strobe signals which synchronize the activities of the different components.
We build a tools such as the SWAS scheduler or the SWAS timeline
generator by selecting different components and plugging them into the
core data structure. Components include a Target object which manages
the SWAS target list and waypoint and gyrocal and planetary targets, an
Orbit object which can either read an existing ephemeris or generate a model
orbit, an ACS object which embodies the state of the spacecraft Attitude
Control System and a Receiver object which manages the science instrument.
A web site, http://www.quaternion.harvard.edu, is under
construction to describe the SWAS planning and scheduling code in
greater detail and to make it available for use by other teams.
© Copyright 1999 Astronomical Society of the Pacific, 390 Ashton Avenue, San Francisco, California 94112, USA
Next: Recent Improvements to HST Parallel Scheduling
Up: Observatory Planning and Scheduling
Previous: TransVERSE: An Architecture for Configuring Astronomical Observations
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint
adass@ncsa.uiuc.edu