A Scalable Embodied Intelligence Platform for Seamless Real-to-Sim-to-Real Transfer of Household Mobile Manipulation Tasks

2026-06-16

AI Summary

This work addresses three key challenges in real-to-sim-to-real closed-loop transfer for household mobile manipulation tasks: the high cost of high-fidelity scene reconstruction, complex policy evaluation, and incompatibility with real-world deployment. To overcome these issues, the authors propose BestMan, a scalable platform that integrates automated scene generation (ASG), a simulation-guided framework for task formalization and skill learning, and a hardware-agnostic unified middleware (HUM) to enable efficient development, seamless integration, and robust cross-domain transfer. The platform establishes a standardized benchmark and demonstrates superior sim-to-real transfer performance across multiple heterogeneous robotic systems, significantly advancing research in household mobile manipulation.

Abstract

Mobile manipulation is a fundamental capability in embodied intelligence robotics. The growing demand for robust and generalizable manipulation in unstructured household environments has driven rapid progress in embodied intelligence platforms. However, achieving a seamless transfer across the real-to-sim-to-real cycle faces three key challenges, including costly high-fidelity simulation scenes reconstruction, the complexity of systematic strategy evaluation in simulation, and incompatible real-world deployments. To address these challenges, we develop BestMan, a scalable and seamless real-to-sim-to-real platform that bridges the gap between the simulation and the real world, enabling effective strategy development, integration, and deployment for household mobile manipulation. Specifically, we design a novel Automated Scene Generation (ASG) module to reconstruct realistic simulations from real observations. Then, we propose a simulation-guided task formalization and skill learning architecture that supports the flexible integration and large-scale evaluations of hybrid skill strategies in simulation. Finally, to enhance the real-world scalability, we develop a Hardware-agnostic and Unified Middleware (HUM) to ensure seamless and compatible sim-to-real transfer across heterogeneous mobile manipulators for real deployments. Experimental results demonstrate the superior performance of our proposed platform in establishing standardized benchmarks and facilitating promising research in the field of mobile manipulation.

Problem

• mobile manipulation
• real-to-sim-to-real transfer
• embodied intelligence
• simulation-to-reality gap
• household robotics

Innovation

• real-to-sim-to-real transfer
• automated scene generation
• simulation-guided skill learning
• hardware-agnostic middleware
• mobile manipulation