AgentWorld: An Interactive Simulation Platform for Scene Construction and Mobile Robotic Manipulation

TL;DR

AgentWorld is a comprehensive interactive simulation platform that facilitates the development and benchmarking of household mobile manipulation skills for robots, supporting diverse tasks and enabling effective sim-to-real transfer.

Contribution

We introduce AgentWorld, a novel simulation platform with automated scene construction and a diverse dataset for training and benchmarking robotic manipulation in home environments.

Findings

Effective sim-to-real transfer demonstrated across multiple imitation learning methods.

Diverse dataset covering primitive and multistage household tasks.

Benchmarking results highlight the platform's utility for robotic skill acquisition.

Abstract

We introduce AgentWorld, an interactive simulation platform for developing household mobile manipulation capabilities. Our platform combines automated scene construction that encompasses layout generation, semantic asset placement, visual material configuration, and physics simulation, with a dual-mode teleoperation system supporting both wheeled bases and humanoid locomotion policies for data collection. The resulting AgentWorld Dataset captures diverse tasks ranging from primitive actions (pick-and-place, push-pull, etc.) to multistage activities (serve drinks, heat up food, etc.) across living rooms, bedrooms, and kitchens. Through extensive benchmarking of imitation learning methods including behavior cloning, action chunking transformers, diffusion policies, and vision-language-action models, we demonstrate the dataset's effectiveness for sim-to-real transfer. The integrated system provides a comprehensive solution for scalable robotic skill acquisition in complex home environments, bridging the gap between simulation-based training and real-world deployment. The code, datasets will be available at https://yizhengzhang1.github.io/agent_world/