SymbioSim: Human-in-the-loop Simulation Platform for Bidirectional Continuing Learning in Human-Robot Interaction

• DOI: 10.48550/arxiv.2502.07358
• Publication Date: 2025-02-11
• Authors: Hao Chen; Yiteng Xu; Yi Ren; Yaoqin Ye; Xinran Li; Ning Ding; Peishan Cong; Ziyi Wang; Bushi Liu; Yuhan Chen; Z. L. Dou; Xiaokun Leng; Manyi Li; Yuexin Ma; Changhe Tu
• Keywords: Human-in-the-loop; Human–robot interaction; Human interaction

The development of intelligent robots seeks to seamlessly integrate them into the human world, providing assistance and companionship in daily life work, with ultimate goal achieving human-robot symbiosis. To realize this vision, must continuously learn evolve through consistent interaction collaboration humans, while humans need gradually develop an understanding trust shared experiences. However, training testing algorithms directly on physical involve substantial costs safety risks. Moreover, current robotic simulators fail support real participation, limiting their ability provide authentic experiences gather valuable feedback. In paper, we introduce SymbioSim, a novel human-in-the-loop simulation platform designed enable safe efficient development, evaluation, optimization interactions. By leveraging carefully system architecture modules, SymbioSim delivers natural realistic experience, facilitating bidirectional continuous learning adaptation for both robots. Extensive experiments user studies demonstrate platform's promising performance highlight its potential significantly advance research