Problem Statement
Overview
Human-robotic teaming is a central component of President Bush's vision for space exploration. Requirements for exploration will continue to evolve as robotic exploration missions face time and fiscal constraints. To meet these changing requirements, further investigation of human-robotic teaming and development of systems that support these requirements is necessary.
The Ames Human-Computer Interaction Group sponsored a Carnegie Mellon Masters of Human-Computer Interaction (MHCI) capstone project to research, design and prototype an interface for robotic planning.
Problem description
The current Mars missions, with the robotic rovers Spirit and Opportunity, require 40 person-hours of work for each day of operation. In addition, for the first few months of the mission, scientists moved to California so that they could work together effectively; while they have now moved back home, they still work together, albeit in an ad-hoc, evolving way.
Future mission scenarios call for smaller, more self-sufficient teams of astronauts controlling a larger set of robotic agents. In order to support rapid daily planning, the number of person-hours must be improved by at least an order of magnitude over current practices and tools. To achieve this, planning must become more efficient. This project will explore a set of robotic activity planning abstractions to assist with the design of effective representations of robotic capability, efficient interactive methods for creating and modifying plans, and clear visualization of the plan's products.
To this end, NASA's HCI Group has specified a simulation that incorporates many of the design challenges of Mars-rover software, yet operates on Earth with readily available tools. While current mission scientists direct the Mars rovers to find and investigate targets of scientific interest, particularly habitats that might harbor, or once have harbored, life. Our project will replace scientific targets with "points": the more points the users earn, the more successful was our software in assisting them to do so. This enables us to focus on the design of software addressing the process of scientific collaboration, not the domain-specific details of the actual scientific tasks, and thus, hopefully, offer more effective suggestions to the NASA HCI group.
Approach
Focus
Recognizing the limitations of our time and our experience, NASA provided us with a choice of several focuses for this project. Based on our interviews, contextual inquiries and other research we decided to focus on two key focuses:
- Collaborative/distributed planning: Effective collaboration among scientists distributed around the world
- Multi-day planning: Specifying multi-day plans for one or more robots
Based on our selected foci, the simulation provided by NASA's HCI Group was modified in order to make use of two participants.
How we approached the problem
With a total of 8 months to work on this project, we split the time roughly into 3 pieces. The first part was dedicated to researching the actual problems that NASA Scientists and Engineers have with the current Martian Rovers. The second part was a short period of idea generation, followed by a longer development phase.
Description of the Simulation
The simulation is designed to capture key elements of robotic planning for Mars Surface Missions.
- Limited Downlink Capability
- Long Term Mission Planning
- Rover Control
- Correction for Navigation Error
- Multi Person Rover Control & planning
- Emulation of Scientific Discover