MORAL
MORAL: A Body-Thinking Robot Foundation Model

Project Description

Current vision-language-action approaches are dominated by symbolic information such as text, lacking reasoning grounded in body structure or internal states. We propose MORAL as a concept for body-centered robot intelligence, placing bodily representations—including morphology—at the core while integrating modalities such as auditory sensing. This concept envisions embodied reasoning and action generation beyond symbols, aiming for future capabilities such as handling unseen situations, adapting force and motion, and generalizing across different morphologies, with the goal of enabling robots to flexibly operate in diverse real-world tasks in manufacturing.

PI

Masaki Murooka

Period

2025.10 - 2027.3

Funded by

JST CREST