Introduction: Expedition Personnel & Support

Rotation 38, April 3 - 17, 2005, at the Mars Society's Desert Research Station was completely dedicated to the field test of NASA's Mobile Agents EVA communications system.

This project was funded in part by a NASA Research Announcement (NRA) through the Exploration Systems Research and Technology program, John Mankins selecting officer. Project members are civil servants and contractors at NASA-Ames at Moffett Field, CA and Johnson Space Center in Houston, TX.

In addition to the 15 NASA employees, people working at MDRS included two SUNY Buffalo geology graduate students, nine visiting colleagues, a part-time engineer, a reporter, and two researchers from Foster-Miller, for a total of 30 local participants. Don Foutz at the Whispering Sands provided critical logistic support, especially for these shakedown two weeks with the diesel generator.

About a dozen off-site scientists constituted and supported the Remote Science Team, led by Shannon Rupert (please refer to the organization chart at the Mission Info page). Other support behind the scenes was provided by NASA Ames Public Affairs, the RIACS administration at Ames, and of course the Mars Society volunteers who provided Mission Support (including especially Tony Muscatello and Julie Edwards).

Special thanks are due to Abby Semple and Brent Garry, who accommodated our schedule to participate, and endured the delays and unpredictability in using an emerging technology.

Objective and Accomplishments:

Our objectives this year were to demonstrate more improvised EVAs with the Mobile Agents system and coordination with two robotic systems:

• Adding to the EVA Robotic Assistant (ERA) robot an arm with shovel and temperature sensor.
• Completely redesigned robot architecture to allow managing multiple tasks in parallel (e.g., join me, watch me, take a panorama, sample here).
• Addition of a second robot (Thibodeaux) that might go 30 mph (and could be reconfigured to carry an astronaut).
• Completely reconfigured agent system for efficiency (e.g., location agent specializes in location name sharing and navigation responses)
• Improvements to voice commanding, including:
  • interacting with the robots to determine their status ("Whom are you following?")
  • "joining" with a robot (so it serves as a relay)
  • astronaut naming of work stations (1 m areas) within work sites
  • requesting locations and navigation information
• Giving more responsibility to the remote science team for setting science objectives
• Routine use of planning tools for crew-RST communications, so EVAs were planned during the rotation, rather than being scripted.
• Automatic mapping of named places, photo locations, sample bags and voice notes onto dynamically stitched TerraServer maps, stored in the ScienceOrganizer database.

We recognized that we would not get everything working this month, primarily because the new additions are ambitious (making this more like 2003 in shaking down the interactions), but also because of funding delays (that prevented buying the relays, computers, and robot parts until March). In addition, some dual-use equipment was snared at the last minute for other projects (including our trucks!), and an alternative LAN and satellite system had to be deployed (as Glenn Research Center lacked funding). Special credit goes to Bill Notley, Ames Emergency Communications, for providing networking and satellite support.

Our daily reports log the progress of the past two weeks. We were unable to complete assembly of Thibodeaux or the arm on Boudreaux, because of parts burning out that were not stocked by vendors. Also, the extensive changes to Boudreaux (and the absence of the lead robot navigation engineer because of her pregnancy), reduced robot operations to teleoperation (though using video frames broadcast to an operator in MDRS). On the plus side, the integration of the new robot architecture into the reconfigured agent system was demonstrated to work in dozens of tests around MDRS.

Probably the most important accomplishment was demonstrating that the Mobile Agents system indeed can now be used for EVAs of a dozen activities (covering on average 2 hours) planned on the same day by the crew (with advice from the RST's review of previous work). Furthermore, the astronauts could be left alone to work for long periods (20- 40 minutes) without prompting by the support team, or without their asking for guidance of what to do or where to go (this information is now provided by the agent system: Where am I? What is the next activity? Where is work site 5?). We also noted that the HabCom person in MDRS supervising system operations has much less work to do than before. That's good, as our goal is to eliminate that function entirely (the notion of "automating CapCom"). The system is now also remarkably efficient (with nearly immediate response to voice commands), though latency depends on computer network bandwidth (which we discovered still requires careful configuration, even with omnidirectional antennas).

With the robots unable to provide active LAN relay support, we deployed instead Tropos repeaters by teleoperating a robot towing a relay on a trailer. We also demonstrated how a person operating an ATV could simulate a robot, which we called the ATV/ERA.

The ATV/ERA is an ordinary ATV with the ERA's agent system (on a laptop computer) and a button interface on a Casio computer mounted up front. The interface computer is connected to the ERA computer by VNC login to a program simulating the ERA executor, which is also running on the ERA computer. Thus when a voice command from the astronaut states, "Thibodeaux, follow me," the agent system on the ATV/ERA eventually receives the command and translates it as something the robot can execute. This information is received by the simulator and presented on the button display for the person operating the ATV to carry out.

The ATV/ERA system allowed us to provide the functionality required (dynamic relay), plus to test the new "join" voice commands that made a robot move to keep the astronaut on the computer network. This system will be useful at Ames in California for further development and testing, as the robots themselves reside at Johnson Space Center in Houston. So when you see a photo of a person on an ATV following the astronauts in our field experiments, you should now realize that a very complex system is actually being exercised. When Thibodeaux has been completely assembled, we will simply substitute the real robot executor for the simulator, and the robot should be behave like the ATV/ERA.

During our rotation, several colleagues visited us from Ames for about a week each. We believe we now understand each other's methods and capabilities far better, and have quickly hatched new plans for working together. (It's amazing what being stuck together in the hab during a rain storm can do for morale.) This work includes integrating the Mobile Agents system with "autonomous" robots at Ames; planning systems that could generalize how the robots respond to multiple, sometimes conflicting requests from different people; and tools to sketch EVA plans that will compile into formal models interpretable by the Mobile Agents system.

We completed four EVAs (Friday, Tuesday, Thursday, Friday), with several trials of "autonomous scouting" by Boudreaux (which failed, but revealed how the new multi-threaded robot system needed to work). In some respects, this was disappointing compared to last year's cinematic EVAs. However, we all understood that everyone had done their best. We worked as quickly as possible over the past six months, and even more intensely during 10-14 hour days at MDRS. Everyone knew that the robot team at JSC was short-handed and underfunded, and that rebuilding the hardware and software was unlikely to be ready by April (it is too warm for the computers to work here much later). But we also know that these two weeks of integration in the field serves us better than a hundred telecons or a few short trips to Ames or JSC. After two weeks, the list of improvements and detailed lessons we have learned would require many pages. Depending on funding we intend to experiment with the same configuration next year.

So fifteen people leave MDRS this weekend with realistic contentment, and a bit exhausted.

Bill Clancey
Chief Scientist, Human-Centered Computing
NASA Ames Research Center

(The opinions expressed here are entirely those of the author, not NASA or its organizations; following are some comments by the Project Lead)

---

Crew 38 Experience
By Maarten Sierhuis

Everyone who has been out to the MDRS immediately feels one with its environment. The beauty, silence (except for the noisy generators), and the mere fact that city life is oh so far away, makes a two week stay at the MDRS an unforgettable experience.

The Mobile Agents teams arrived in Hanskville on Friday the 1st. I had to fly in on April 2nd due to the fact I had to teach my class the day before. I couldn't easily skip three weeks of classes, even though my TA was there to take over. Because of this I missed the Crew 37 handover and when I arrived at the upper deck, and put my bags in my recognizable stateroom, I felt a little out of place. Much of the hardware, wiring, electricity and water systems had changed from the previous year. Without getting that introduction, it took me about two days to become familiar with the new systems. The one good thing about being on a NASA team, is that almost everyone is an engineer (and a smart one!), and thus I had about 15 people I could ask how something worked.

This is the third year in a row that the Mobile Agents team has its field experiments at the MDRS. We have thus formed a sort of work practice that people fall into as soon as they arrive at the hab. One thing that I noticed immediately is that this year, almost no one had to be told where to sit, what to do, where to install their systems, etcetera. One could model the work practice at the MDRS of the Mobile Agents team. You could first start with specifying the teams and roles:

Mars Society Members Mission Support Team Flight Surgeon Mobile Agent Team Ames Team Brahms Team MEX Team MEX System Team NREN Team RIALIST Team Science Organizer Team JSC Team ERA Team University of Buffalo Team UB Geologist Mobile Agents Team Lead Mobile Agents Prinipal Investigator Mobile Agents Project Lead Brahms Team Lead MEX Team Lead RIALIST Team Lead JSC Team Lead Remote Science Team Lead

UK Teams KMi Open University Compendium Team KMi WebLog Team Southampton University Science Media Player Team MDRS Crew Member Commander Health and Safety Officer Director of Galley Operations Engineer Crew Engineer Mission Support Engineer Local MDRS Support Remote Science Team RST Geologist RST Biologist RST Back Room RST Facilitator Visitor NASA Visitor NASA Collaborators Media Visitor School Visitor

This is just a start and I most likely have forgotten a team or two, but you can clearly see the complexity of the Mobile Agents project. Imagine the complexity of modeling the practice of this group of people. Where are they all located, how do they interact, when, where, for what purpose, etcetera? People create incredible complex structures, but are able to fluently live within this complex social network. How do robots fit in? Can they ever understand, or even just deal with this kind of social network? We have our doubts. This year the robots played a lesser role. Not by choice, but by circumstance.

We have learned much again. We will surely improve all our systems. I personally feel we're getting closer to a real workable system. The system is more robust, faster, more capable and easier to use. However, it will take more work to make it usable for an average MDRS crewmember, without our involvement. That would be nice, wouldn't it? Provide every MDRS crewmember with his or her own Mobile Agent System when going on an EVA. I secretly have this as a goal.

Till next year!!! Hopefully!!!