Multi-agent Planning using Visual Language Models

ECAI 2024

Put a clean tomato on the counter (trial_T20190909_012550_586494)

Ground truth Single Agent with table Multi Agent with table Single Agent with image Multi Agent with image
STEP 1: Walk forward then turn left and walk to the fridge.
STEP 2: Open the fridge and grab the tomato by the mug then close the door.
STEP 3: Turn left and walk to the sink then take a big step left.
STEP 4: Wash the tomato in the left sink and then pick it back up and shut off the water.
STEP 5: Take a step to your right.
STEP 6: Put the tomato under the sponge to your right.
Turn around and walk to the fridge.,
Take the tomato out of the fridge.,
Turn around and walk to the sink.,
Wash the tomato in the sink.,
Turn around and walk to the counter.,
Place the clean tomato on the counter.
Turn around and walk to the sink.,
Take the tomato from the sink.,
Turn on the faucet and wash the tomato.,
Turn off the faucet and shake off excess water from the tomato.,
Turn around and walk to the counter.,
Place the clean tomato on the counter.
1. Walk to the sink located on the counter.
2. Pick up the tomato from the sink.
3. Turn to face the counter space to the right of the sink.
4. Place the clean tomato on the counter space to the right of the sink.
1. Walk to the sink, which is located under the window.
2. Pick up the red tomato, which is on the counter to the right of the sink, between the green apple and the brown onion.
3. Turn on the faucet, which is above the sink, to get water.
4. Wash the tomato under the running water from the faucet.
5. Turn off the faucet.
6. Place the clean tomato on a clear area of the counter, to the right of the green cup.
Architecture