notesum.ai

Published at November 11

Enhancing Robot Assistive Behaviour with Reinforcement Learning and Theory of Mind

cs.RO

cs.AI

cs.HC

Released Date: November 11, 2024

Authors: Antonio Andriella¹, Giovanni Falcone², Silvia Rossi²

Aff.: ¹Artificial Intelligence Research Institute (IIIA-CSIC), Barcelona, 08193, Spain; ²Department of Electrical Engineering and Information Technologies, University of Naples Federico II, Naples, 08005, Italy

Arxiv: http://arxiv.org/abs/2411.07003v1

Action

Card

Inferred information

Implementation Example

1^{st}

Both locations have been visited once

“You have seen both locations of shark once. Let me refresh your

memory: row 1 and col 2.”

sug_card

1^{st}

One location never visited, the other visited multiple times

“You have clicked shark several times. Click the one in row 1 and

col 2, you should then remember where the other one is located.”

1^{st}

Both location visited multiple times

“You have often seen both locations of shark. The one less visited

is located in row 1 and col 2. In this way, you should make a match!”

1^{st}

One location visited multiple times, the other has never been visited

“Are you looking for a particular card? Well, then try row 1 and col

2. Surely you remember the other location!”

2^{nd}

Both locations have been visited once

“You’ve seen this card before. I remind you that it is located in row 1.”

sug_row

2^{nd}

One location clicked multiple times, the other has never been visited

“This card again? You’re struggling on this one. Well, then try row 1.”

2^{nd}

Both locations visited multiple times

“You have seen both locations of shark more than once. Try to

remember at what location of row 1 the other card is located.”

2^{nd}

Current location only visited once, the other has never been visited

“You haven’t seen a shark before, so let me help you: try the third

row.”