|
|
Acting as robotic geologists, two six-wheeled rovers are scheduled to land
in early 2004 on opposite sides of Mars to search for evidence of water
or a history of water on the planet. Each Mars
Exploration Rover's planned mission lifetime is ninety sols (martian
days) and it can perform several hours worth of activity per day (e.g.,
taking panoramic images or driving). Each martian
day, commands will be uploaded to the rovers. Due to limited data capacity
and significant lag between command issuance and data return, the science
discovery and planning process is highly time-pressured. The mission science
and engineering teams have 18 earth hours to analyze the data received,
use the analysis in deciding the science plan for the following day, and
encode the science activities as a sequence of commands for transmission
to the rover.
Having analyzed the previous day's data and created a set of high level
requests for activities to be done by the rover the following day, the scientists
go off shift as the engineering team is coming on shift. The engineers need
to generate a sequence of commands that are consistent with scientific intent
in order for that data collected to be scientifically useful. For example,
three images must be done an hour apart to measure atmospheric change over
time. Without all three images in that time frame, the data will not be
useful to the scientists.
The lead scientist for the day works with a mission engineer to encode
scientific intent using a tool called Constraint Editor, for example,
specifying that one activity must come before or after another or at a
particular time of day. These constraints are strictly enforced by a scheduling
tool called MAPGEN that
the engineers use later in the process. However, entering constraints
without omissions or errors is very difficult. The MAPGEN
team asked the Human-Computer Interaction (HCI) team in the Human Factors
Research and Technology Division at NASA Ames to design the Constraint
Editor to support the efficient and accurate capture of scientific intent
through the mission critical uplink process.
The HCI team worked with lead mission scientists and engineers to design
Constraint Editor. As the HCI team understood the problem, they realized
scientists could not keep the large number of constraints (often more
than 100) in their heads. Successful execution of the task required a
representation of constrained activities, activities not yet constrained,
and a real-time view of the state and correctness of the developing plan.
The HCI team iterated with JPL through several prototypes of the tool.
Concurrently, the HCI team provided mission scientists and engineers training
on the tool and concept of constraint-based planning. The HCI team worked
with JPL to transform the mindset of experienced scientists and engineers
from the familiar method of hand sequenced plans to a method involving
automated constraint satisfaction. The combination of the HCI design and
focused training has led to the successful adoption of constraint-based
planning and of the tool for the Mars
Exploration Rover's 2003 mission.
|
|