FOCUS: Object-Centric World Models for Robotics Manipulation

arXiv 23.7 2023 55.8 method, application

TLDR

FOCUS learns an object-centric world model for robotics manipulation, using exploration bonus to improve task efficiency and real-world deployment.

Reasoning

The paper introduces a novel object-centric world model with an exploration bonus, demonstrating improved task efficiency and real-world applicability. Strengths include clear focus on structured representations and empirical validation with a robot arm; weaknesses are limited detail on baselines and broader comparisons in the abstract.

Read-first score

Read-first score 55.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 33.

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=True; dataset=False; markers=github

Recency 8%

65.1

Uses a gentle age decay so recent papers surface without erasing older foundations. 2023

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=experiment

Topical relevance 42%

47.1

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

Reproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 189.

Keyword Scores

world model

model-based reinforcement learning world model

world dynamics prediction

interactive world model

generative world model

world simulator

video world model

Deep Analysis

Innovations

Proposes FOCUS, a model-based agent that learns an object-centric world model for robotics manipulation.
Introduces a novel exploration bonus derived from the object-centric representation to encourage exploration of robot-object interactions.
Demonstrates that object-centric world models enable more efficient task solving and consistent exploration across manipulation tasks.

Methodology

FOCUS is a model-based agent that learns an object-centric world model. It uses a novel exploration bonus stemming from the object-centric representation to facilitate exploration of robot-object interactions. The approach is evaluated on manipulation tasks across different settings, including real-world experiments with a Franka Emika robot arm.

Key Results

Object-centric world models allow the agent to solve tasks more efficiently and enable consistent exploration of robot-object interactions, as shown in both simulated and real-world settings.

Resources

Paper Code