Behavioral Knowledge That Different Robots Can Share

The options

The order shown is not a ranking of preference.

1. Option 1

   Share movement trajectories

   Reuse the accumulated successful movements and practice by converting them for each body. Initial observation design and environmental change are minimal, but conversion and re-learning are needed each time a body is added, and records about material change grow thin.

2. Option 2

   Share object change and feedback

   Share how the paper changed after the action and what was observed to make adjustments. There is room to find movement anew for each body, but observation, interpretation, and exploratory trial costs are high, so improvements to the common environment must be reduced.

3. Option 3

   Share the common work environment

   Co-design a work environment that helps with paper layout and contact, and have each body practice within it. There is room to reduce differences in performance, but modification, maintenance costs, and re-layout time are incurred, and adaptation practice outside the environment decreases.

The situation as it stands

If the commitment is not set, the repository and workbench build schedules will each slip, the secured joint trial period will be spent on foundation preparation, and learning by body will accumulate in states that are hard to compare with each other.

The low-risk paper-model workshop of the fictional Folding Tabletop Experiment Coalition

1. Premise 1Synthetic premise: low-risk tabletop robots of differing size and joint configuration fold lightweight paper models. Even when a successful movement trajectory from one robot is applied to fit another body, the folded form differs, so the procurement of a shared learning foundation becomes an issue.
2. Premise 2Synthetic premise: the budget covers only one main shared foundation and comparison trials. Trajectory conversion requires re-learning by body, change records require observation and interpretation time, and a shared environment requires workbench modification and re-layout time; none of these can be expanded.
3. Premise 3Synthetic premise: the existing robot operations team wants to use the accumulated trajectories, and the new body design team wants to share the regulations that follow folding changes. The workbench management team proposes matching the environment but must take on the modification schedule and the maintenance burden.

Where it holdsChoose one main shared unit to build with this joint learning budget.

What can be lost

• The existing robot operations team may lose the chance to reuse its trajectory assets, or lose time to the conversions that must be performed for each new body.
• The new body design team may lose its own exploration time or the chance to test a body that does not fit the existing foundation.
• The workbench management team may lose resources for modification and maintenance, or the operational room to keep a range of layouts open.

What is still unknown

1. Unknown 1How much of the failure remaining after conversion by body comes from paper condition and contact differences is unconfirmed.
2. Unknown 2Whether, from the same paper change records, each robot can find the next action it can execute is unknown.
3. Unknown 3Whether success in the common environment will persist after part of the workbench has changed has not yet been observed.

Signals to decide again

• The experiment coordinator tallies, each time a body is added, the re-learning time until that body's first success, and reopens the decision if the time saved by reusing shared materials does not decrease.
• The comparison observation team checks each month the differences in possible subsequent actions after the same object change, and reopens the decision if cases accumulate that cannot be compared from records alone.
• The facility keeper checks, each time the workbench layout changes, the change in performance and the recovery cost, and reopens the decision if environment-dependent cost exceeds the agreed scope.