Limiting the Focus

Overview

When the project began, the client gave us a choice of four possible foci:

  • Collaborative/distributed planning
  • Multi-day planning
  • Multi-agent planning
  • Contingency planning

Each area is of interest to the scientists and engineers on NASA's current and upcoming Mars missions. The client acknowledged that undertaking all four foci would be too much for a school project, and invited us to choose the ones we liked, based on our interests and the information we found during our research.

During our research, we moved gradually toward a practical understanding of each item of focus, and eventually chose to limit our focus to “collaborative/distributed planning” and “multi-day planning.” However, as we began designing solutions, and testing those designs, two more real-life needs arose:

  • Rover planning (in general)
  • Interpreting and correcting for rover error

It turned out that we needed to address these two issues before even beginning to tackle the first four foci. Thus, these two issues became the core of our design and development efforts. We never abandoned the two foci we'd chosen from the client's list, but we set these two as higher priorities.

Here are the reasons we chose or rejected each focus.

 

1. Rover Planning (in general)

Rover planning refers to the process of creating a set of instructions for a robot, transmitting them to the robot, retrieving the results after a certain time delay, and updating the user's display based on each result.

This is one of our client's current responsibilities for their customers. As such, the client was very interested in seeing our interpretations of how to address this problem. This meant that we needed to create an infrastructure enabling a user to specify a plan for a robot at all, before enhancing that plan with features suggested by the other foci.

 

2. Interpreting and Correcting for Rover Error

Interpreting and correcting for rover error refers to the fact that the rover doesn't go where you tell it to go; it drifts. On Earth, this is because our mini-rovers' wheels aren't quite accurate; on Mars, it's because the rovers are driving across sand, up and down hills, and so on.

This led to two significant complications in our design and development efforts. First, we had to clearly communicate the difference between the commands the user intended the robot to accomplish and the commands the rover actually performed. Second, we had to enable the customer to adjust for those differences. If we couldn't do this, we couldn't do any of the more advanced, "interesting" topics below. However, this situation turned out to be a fairly advanced, interesting issue by itself.

 

3. Collaborative/Distributed Planning

Collaborative/distributed planning refers to geographically distributed people working together on the same project.

This is both a current, practical concern and a pressing foreseeable concern.  As a primary example, the current Mars missions started as 90-day efforts, during which dozens of scientists moved to California and lived on Mars time to perform their work.  Now, two years later, they have returned home. However, the team still meets for 12 hours a day, 5 days a week, staying in contact by e-mail, phone, and videoconference.  NASA needs to reduce the cost and time associated with helping these scientists collaborate, by enabling them to collaborate efficiently and effectively while staying in their home locations. It is an issue that recurred repeatedly during our research.

 

4. Multi-Day Planning

Multi-day planning refers to specifying a plan for the robot that spans several days, enabling the robot to be more efficient with its time, and allowing the humans to take more practical breaks (such as going home at night).

This is also a current and predicted issue.  The current missions are only active during the Martian daytime.  The mission team typically sends each rover one day's activity at a time, meeting each day to review the previous day's downloads and plan the following day's activities. However, the Mars Science Laboratory, expected to come on-line in 2009, will have its own nuclear-power generator, able to perform science without pause for days or weeks (or years). To take advantage of this, NASA needs to be able to specify multiple days’ activities at a time. In fact, this already happens; today's mission scientists “go home” for the weekend, and thus send the robot a 2- or 3-day plan every weekend. However, it will be an even more significant issue in the future.

 

5. Multi-Agent Planning

Multi-agent planning refers to situations involving teams of robots, satellites, and/or people.

While we know of several research projects involving teams of robots, and we interviewed a number of people on those projects, there are currently no planned NASA missions requiring this area of focus.  While we remained alert to situations that suggested practical, near-term usage of multiple robotic agents, none arose during our 8 months on the project.

 

6. Contingency/Branching Planning

Contingency/branching planning refers to mission scientists' ability to specify multiple possible plans of action for a robot, and enabling the robot to choose one of those plans.

Most of the contingency planning encountered in current Mars-like robotic missions seems to be human, not robotic.  For example, on NASA’s current Mars Exploration Rover missions, a robot will stop operations when it encounters a problem, waiting for a human to send it instructions the next day.  Similarly, on the CMU/NASA project Life in the Atacama, the robot may additionally choose to bypass particular a science target if it encounters a problem—leaving scientists to discover, after the fact, why it chose not to accomplish a given task. NASA has informed us that, for the foreseeable future, robots will not have enough on-board processing power to support advanced scientific decision-making. This means that each contingency comes back to the humans on a project as a new set of decisions to make. In other words, contingency planning is handled by a manual, human, social process, which seems to suffice for current and predicted situations.

 

Summary

Multi-day planning and collaborative/distributed planning are current and predictable needs for Mars missions; as such, we chose these as our primary foci.

However, most of our time was spent designing and building a system for creating rover plans and adjusting for rover error.