The Robot Stuff
Swarm
- That’s interesting.
- What would make it not interesting anymore?
- Even if a study comes along just changing parameters on top of that, it won’t be interesting.
- The discovery that the sense of body ownership works is interesting.
- It feels like they are expanding the “body” model.
Maybe expanding the model is what makes it interesting?
High-level async already exists, and I even got a paper through, so it’s not interesting anymore and lacks novelty. If you can come up with even more amazing new concepts on top of that, that would be interesting. Simply comparing new visualizations won’t make it a hit paper unless it has abstract breadth and impact itself.
For example,
- Going high-level and reversing time?
- Moving between multiple spaces and multiple times?
If you delve in one direction, it might just end up being a plain CAD, or maybe CAD + timeline.
- I think Steve might be imagining that.
- But if that happens,
For example, applying the paradigm of asymmetry again. The question of “what to share”
- Let’s assume robots are distant players.
- The instructions are what they want to share, but optimization in space is individual.
- Well, there’s no individual optimization on the system side.
- It only uses all the data.
- But maybe the system side is just a normal environment. Different ways of presenting can be possible when used by person A and person B.
- That’s what I want to explore. Just asymmetrical space is,
- What can’t be done without high-level?
- Asymmetry.
Distorting spatially but still maintaining the arrangement?
- There are many blocks lined up, and the user gives instructions, but they are transformed and conveyed to the robot (or human).
- Can this be done from an FMRG perspective..?
I’m not very interested in creating complex systems. I want to implement simple and straightforward concepts.
As a straightforward idea,
Hypothesis
- Can it be done quickly by arranging optimally?
- Different ideas come up?
- Can you see from a perspective that is normally invisible?
- Can you switch the arrangement?
Big hypothesis
- In an asymmetrical way to real space.
Placing everything within reach by distorting the coordinate system.
It’s continuous and tough if it’s not high-level. It works with limited information.
Relative distance.
If you want to make high level even higher, it’s about relative distance.
Highlighting only strong relationships and diminishing everything else.
More generalization
- Inferring high-level intentions from actions. What’s the hypothesis?
- Can you move objects by overlapping multiple spaces?
What I’m doing seems similar to overlap.
- Is the novelty in being able to group and move?
Indirect manipulation
Purpose?
- Manipulate objects quickly.
- Teleportation or something like what ritar-san does might be good.
- I feel like I’ve seen something like what ritar-san does somewhere.
- Can’t think of any other way to evaluate.
- As for the asymmetry philosophy, if individual optimization brings happiness, then anything is OK.
- It’s important to evaluate if solving the asymmetry issue arising from individual optimization is possible.
Technically useful stuff
- Providing guides when placing.
- There seem to be various things in 2D.
- https://www.sciencedirect.com/science/article/abs/pii/S0950584915001718
- Relative position system
- yyyggg