Exploration

Exploration Out of Focus

Robotics

Literary Review

Since the goal of our project is to design an application/tool to support planning for Mars rovers, it was only obvious that we should look at robotics papers and projects around Carnegie Mellon University. Literature analysis provided the team with data on a variety of topics relevant to robotics. While the majority of this data was not directly related to our foci, we believed that they might still help motivate our final designs.
    The contents of this section include:
    a. Common Grounding
    b. (Semi) Autonomous Robotic Interaction
    c. General Human Robotic Interaction

Common Grounding

The papers described projects that have people from different backgrounds and sometimes in different locations require that each person have a common understanding of the other members of the team. The papers and research in this area focus on the importance of common understanding, challenges to grounding and in some cases offer solutions to this problem.

Insights:

Common grounding is required for teams to work effectively in distributed environments and it is important to communicate the goals and reasons behind each group’s actions.
The team’s success depends on the ability of each team member to understand the perspective of other members. The most effective human-robot teams will be those that utilize a collaborative model of the environment and task.
It is sometimes assumed that [robot] autonomy is the ultimate goal for robotic systems. One purpose of the research reported is to suggest that effective teamwork, where the robot is a [trusted] peer, is an equally profitable aim.

(Semi) Autonomous Human Robot Interaction

Robots can be fully automatic, or can require various amounts of human attention. The papers in this section describe how humans can interact with autonomous and semi autonomous robots. The paper discusses how a robot can be used as a museum tour guide. The paper focused on explaining that museum robots must be completely autonomous, with a low breakdown rate. The robot must be able to move around and educate the public, and recharge itself all with minimal to no supervision. Another paper discusses some of the good and bad things done in semi-autonomous robot interfaces.

Insights:

In situations where multiple people must observe information (in crowds) multi-model presentations are helpful.
A large number of errors can be resolved simply by "retrying". Errors like missing the recharge station, or trying to navigate, many of the obstacles or problems could be resolved simply by attempting the operation a second time. If the second time failed, the robot should call for help.
The interface should have displays to help the controller understand the robot's environment and achieve situational awareness; this can be done with the use of both video & map displays, data displays (position, speed, tilt angle, etc.), and 3D models (but these should only be used when highly accurate; otherwise they may do more
damage than good).
Knowing more about the context of the rover's surroundings beyond than the video feed is important.
The interface aspect that allows for the user/controller achieving situational awareness of the rover, whatever method is used should be highly accurate so as not to do more damage than good.
Conveying the rover's autonomy to the user is important and the user should not lose access to any control over the rover's autonomy that is available.

General Human-Robotic Interaction

Included in this section are papers that do not fall directly into the other sections. This includes documents on training, and how people learn about robots. An example topic is the Personal Exploration Rovers (PER)s and the museum exhibits that utilized them.

Insights:

It's important for the UI to accurately represent the robot's environment.
The interface should allow for varying degrees of autonomy.
The operator should have situational awareness of the robot.
One way to convey situational awareness is by showing panoramic views.
PERs' heads tilt and pan as it "looks around," because although that makes the robot's job harder, it was much, much better as a diagnostic tool for the people driving it.
The PER game is analogous to the game which is our target product: "...show that rovers are a tool for doing science by enabling visitors to act as mission scientists, using the PER to conduct a science operation"; "an immersive tool for experiencing the challenges faced by NASA mission scientists."

Interviews and Contextual Inquiries

Robotics professor, researching Robotic Telepresence

The robotics professor we interviewed worked on a project that could give the human a better experience telepresently than in a space suit. The project involved many disciplines: communication, stereo-vision, the robot itself, and the scientific specialists for the current mission. He discussed multiple agent interaction such as multiple robot platforms in the project, and a couple of humans monitoring them and collaboration aspects: such as a multi-tiered control architecture, distributed work such as a person driving the robot, possibly, locally and handing off science request to somebody remotely. He gave much feedback on contingency/branching aspects. One problem that exists is how do you pre-program “experience”? Also, can we have other people on the outside intervene?

Insights

A lot of contingency/branching planning can be handled well by having a bunch of people with lots of experience contribute their ideas. See the “open-source” design idea for more information.

Collaborative Planning

Literary Review

This section focuses on areas not specific to either science or robotics. This area researched how people work collaboratively in teams and how people come up with plans and schedules in various contexts. As such, we looked into research literature in computer supported collaborative work (CSCW) as well as real world situations where people created plans and schedules.

CSCW Papers

Because one of our foci dealt with how a team of people collaboratively worked together to come up with a plan, we looked into the field of CSCW (computer supported collaborative work) for basic insights into group work. topics that we determined relevant included: groups, online communities, coordination, virtual teams, grounding and awareness/overload. Some example papers dealt with the development of the B2 Bomber and a case study that examines how a distributed team was able to complete a large project cutting the production cost from $4.5 million to 47 thousand, while each spending no more than 15% of their time.

Insights:

Having a common grammar: Decide in advance how you communicate between parties, having specific tolerances for the common data.
Grounding is the fact that teams of greater than one people require communication in order for all members to be on the same level of understanding of whatever task they wish to perform.
The use of alternative descriptions will allow for better grounding (strong need for shared vocabulary, and visual design to tie information together) and the use of referential installments breaks concepts into smaller digestible chunks.
We should allow for commenting to facilitate a discussion or conversation to clear up or present possible misunderstandings.
The main benefit of collocation is the continuous reassurance that the team has a shared understanding of the issues at hand.
There is a need for a centralized data view that all people can access in the same way
Maintaining "mutual knowledge" i.e. common ground is a central problem of geographically dispersed collaboration.
Mutual knowledge can be established through interaction, but when a group is mediated through technology, mutual knowledge can be worse.
Visibility of communication is key to having mutual knowledge and goals, incentives, and situations should be aligned whenever possible.
Managing tasks involve issues of negotiating with whom to communicate and under what conditions.

Interviews and Contextual Inquiries

We have initially observed from a few places that do real world planning in order to obtain some insights into other contexts. People already plan with multiple people multiple days in advance for big projects and we wanted to understand how people do this currently.

Game Company

We performed a contextual inquiry at a small fifteen member game company that takes on around five projects at a time. We observed a weekly status report meeting where the company’s CEO discussed the status of the company with every team member present.

One interesting observation was the fact that they made their own planner specifically to assigning tasks. The artifact model (see below) brought out the specifics of what information was most important. The planner tool allowed people to switch around easily, was based on task manipulation, was deadline driven and made commenting and editing easily accessible. While the meeting’s intent was to check up on existing plans, not to create a new plan, we got some insights as to how planning in advance would require such updating. The CEO’s main interests were in making sure that members weren’t overburdened for a particular deadline and that project deadlines were on their way to becoming completed.

Insights

Deadline driven plans have less emphasis on the actual calendar
time
Multiple views of the same data are useful in providing different perspectives especially since multi-day planning introduces greater complexity than just single day
Tasks may have to be clarified in person if updated information keeps coming in too fast to document

Hospital Planning, Information Technology Research (ITR)

We conducted an interview with a Ph.D. student observing the planning and scheduling in hospitals. We asked questions regarding his experience observing anesthesiologists, surgeons, and nurses plan and schedule operations for over a six-month period. He is interested in the role of team member awareness on individual attention allocation choices in multi-project, multi-team collaboration. Additionally, he is exploring shared displays in the Operating Room environment that support inter-team collaboration and coordination. Part of which, he mentioned the use of the central whiteboard. The whiteboard was used to keep track of patient-to- room scheduling and team member tasks.

Insights

Scheduling is like running a playbook
Something as simple as a whiteboard can be an effective only if supported by an organized system
Physical space influences coordination

Science

Interviews and Contextual Inquiries

In order to understand how scientists on MER missions collect data and plan missions, we interviewed local scientists doing similar work. We talked to scientists about planning, data collecting, modeling, and scientist/engineer collaboration. We interviewed an atmospheric chemist with training in geochemistry and meteorology, a geologist currently doing remote work with Mars volcanic structures, and a mineralogist who has had years of practice in expeditions collecting minerals from all around the world.

Atmospheric chemist / Geochemist

We interviewed an atmospheric chemist/organic geochemist. We asked questions about his recent field and lab work. His recent expedition was a mission to investigate Soviet gases by measuring hydrocarbons in the marine atmosphere. Weekly meetings were held to continue hypothesis testing and to discuss objective status updates. The mission lasted for over two years. Particle measuring would span over multiple days. The scientist also runs a research group made of scientists and graduate students. Their most recent expedition involved investigating air toxics at a narrow sliver of land. The team uses a mobile laboratory and some of the county's air monitoring stations to measure and record pollutants. The scientists and researchers are also working to build computer models that can predict the effects of reducing specific sources, predicting how chemistry transforms particles in the atmosphere and predict the influence of those particles on global climate. The objective is to test with in-situ observation the oxidation mechanisms developed in laboratory studies.
Insights:

Scientists work with multi-disciplinary colleagues but handle most of the tasks independently
Instrument tweaking and data gathering processes can span for
many hours if not days at a time

Mineralogist

We interviewed a mineralogist that works as a museum curator in the gems and minerals section. He has six years of exploration as a geologist tracking and finding mineral deposits. He generously shared adventurous stories of his past expeditions such as his search for uranium in the Black Hills of South Dakota. He noted that for each expedition, there is a process that needs to be followed in order to be successful:

Research the topic
Organize and gather resources
Prepare the expedition team
Form a cohesive plan for the project

Insights

Since scientists are constantly looking to collect more data, goals must be made certain.
Scientists need to know why, therefore nothing is generalized
Engineers design how to get around the problem, which causes discrepancies among scientists and engineers.
Maximize the amount of data... collaboratively make informed decisions...distribution of information is key
Time is the biggest factor that is why careful data collection is important
Bull Sessions are free-association, informal discussions- which is crucial (may be interesting in a tele-conference)
Some of the best discoveries are serendipitous and accidental but by trained observers (think outside of the box)
Exchanging information with people who aren’t colleagues is important (Gem & mineral shows)

Remote Input Sensing Geologist

The geologist we interviewed specialized in remote sensing geology. He works with images taken remotely and remote sensing tools such as satellite cameras, and different kinds of camera images such as thermal and heat output images. He shares traditional methodology that involves going into the actual area of rock outcrop and non-traditional methods that are more like archaeology: planning and laying out a grid area for future exploration.

Insights:

Multi-day is inevitably what happens, always have to plan in advance.
Factors that drive decisions: safety precautions, deciding where to go, deciding where to put cameras.
Traditional and non-traditional geologists have different planning and collaboration methods
Science is so broad, how detailed should we be studying?
Tag areas of what you want to get to
The need to get to certain area involves contingency plans