Search papers, labs, and topics across Lattice.
This paper introduces BestMan, a scalable platform designed to facilitate seamless real-to-sim-to-real transfer for household mobile manipulation tasks. By addressing challenges such as high-fidelity simulation reconstruction and systematic strategy evaluation, BestMan integrates an Automated Scene Generation module and a simulation-guided task formalization architecture. Experimental results show that this platform significantly enhances the performance and compatibility of mobile manipulators across various real-world deployments, establishing standardized benchmarks for future research.
BestMan enables seamless integration of real-world observations into simulation, revolutionizing how mobile manipulators learn and adapt in unstructured environments.
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.
