RY

C.03 / ADAPT

2025—2026

TOIO AGENTS

How can multi-agent behavior become observable, physically situated, and open to correction?

Multi-agent toio physical task

A Unity-to-physical testbed connecting multi-agent training, overhead localization, BLE control, and tabletop robots.

ROLE
System and experiment development
STATUS
EXPERIMENT / ACTIVE

COMPLETE PROJECT RECORD

Every available drawing, image, and process record is kept in sequence. Nothing is cropped into a decorative slideshow.

Multi-agent toio physical task
COVER / PHYSICAL TASK

The tabletop task makes multi-agent behavior directly observable.

TOIO simulation and control pipeline
SYSTEM / CONTROL

The diagram separates training, localization, communication, and physical actuation.

TOIO PPO training environment
METHOD / TRAINING

Simulation training is presented as process evidence rather than final transfer performance.

TOIO carrying scenario
SCENARIO / CARRYING

A carrying task records one embodied cooperation scenario.

TOIO performance scenario
SCENARIO / PERFORMANCE

A second scenario tests coordination through visible movement.

TOIO public demo poster frame
RECORD / PUBLIC DEMO

The still frame identifies the public demonstration record.

VIDEO / PUBLIC DEMO

The recording preserves observable physical behavior of the implemented controller.

Embodied control makes a final demand: every machine decision should remain traceable before editing or acting.