Gemini Robotics 1.5: Pushing the Frontier of Generalist Robots with Advanced Embodied Reasoning, Thinking, and Motion Transfer

저자: Gemini Robotics Team, Abbas Abdolmaleki, Saminda Abeyruwan, Joshua Ainslie, Jean-Baptiste Alayrac, Montserrat Gonzalez Arenas, Ashwin Balakrishna, Nathan Batchelor, Alex Bewley, Jeff Bingham, Michael Bloesch, Konstantinos Bousmalis, Philemon Brakel, Anthony Brohan, Thomas Buschmann, Arunkumar Byravan, Serkan Cabi, Ken Caluwaerts, Federico Casarini, Christine Chan, Oscar Chang, London Chappellet-Volpini, Jose Enrique Chen, Xi Chen, Hao-Tien Lewis Chiang, Krzysztof Choromanski, Adrian Collister, David B. D'Ambrosio, Sudeep Dasari, Todor Davchev, Meet Kirankumar Dave, Coline Devin, Norman Di Palo, Tianli Ding, Carl Doersch, Adil Dostmohamed, Yilun Du, Debidatta Dwibedi, Sathish Thoppay Egambaram, Michael Elabd, Tom Erez, Xiaolin Fang, Claudio Fantacci, Cody Fong, Erik Frey, Chuyuan Fu, Ruiqi Gao, Marissa Giustina, Keerthana Gopalakrishnan, Laura Graesser, Oliver Groth, Agrim Gupta, Roland Hafner, Steven Hansen, Leonard Hasenclever, Sam Haves, Nicolas Heess, Brandon Hernaez, Alex Hofer, Jasmine Hsu, Lu Huang, Sandy H. Huang, Atil Iscen, Mithun George Jacob, Deepali Jain, Sally Jesmonth, Abhishek Jindal, Ryan Julian, Dmitry Kalashnikov, M. Emre Karagozler, Stefani Karp, Matija Kecman, J. Chase Kew, Donnie Kim, Frank Kim, Junkyung Kim, Thomas Kipf, Sean Kirmani, Ksenia Konyushkova, Li Yang Ku, Yuheng Kuang, Thomas Lampe, Antoine Laurens, Tuan Anh Le, Isabel Leal, Alex X. Lee, Tsang-Wei Edward Lee, Guy Lever, Jacky Liang, Li-Heng Lin, Fangchen Liu, Shangbang Long, Caden Lu, Sharath Maddineni, Anirudha Majumdar, Kevis-Kokitsi Maninis, Andrew Marmon, Sergio Martinez, Assaf Hurwitz Michaely, Niko Milonopoulos, Joss Moore, Robert Moreno, Michael Neunert, Francesco Nori, Joy Ortiz, Kenneth Oslund, Carolina Parada, Emilio Parisotto, Amaris Paryag, Acorn Pooley, Thomas Power, Alessio Quaglino, Haroon Qureshi, Rajkumar Vasudeva Raju, Helen Ran, Dushyant Rao, Kanishka Rao, Isaac Reid, David Rendleman, Krista Reymann, Miguel Rivas, Francesco Romano, Yulia Rubanova, Peter Pastor Sampedro, Pannag R Sanketi, Dhruv Shah, Mohit Sharma, Kathryn Shea, Mohit Shridhar, Charles Shu, Vikas Sindhwani, Sumeet Singh, Radu Soricut, Rachel Sterneck, Ian Storz, Razvan Surdulescu, Jie Tan, Jonathan Tompson, Saran Tunyasuvunakool, Jake Varley, Grace Vesom, Giulia Vezzani, Maria Bauza Villalonga, Oriol Vinyals, René Wagner, Ayzaan Wahid, Stefan Welker, Paul Wohlhart, Chengda Wu, Markus Wulfmeier, Fei Xia, Ted Xiao, Annie Xie, Jinyu Xie, Peng Xu, Sichun Xu, Ying Xu, Zhuo Xu, Jimmy Yan, Sherry Yang, Skye Yang, Yuxiang Yang, Hiu Hong Yu, Wenhao Yu, Wentao Yuan, Yuan Yuan, Jingwei Zhang, Tingnan Zhang, Zhiyuan Zhang, Allan Zhou, Guangyao Zhou, Yuxiang Zhou | 날짜: 2025-10-02 | URL: https://arxiv.org/abs/2510.03342 📄 PDF

라이선스: CC BY

Essence

Figure 1 | The Gemini Robotics 1.5 family of models consists of Gemini Robotics 1.5, a VLA, and Gemini

Gemini Robotics 1.5는 Motion Transfer 메커니즘과 embodied thinking 능력을 통해 다중 로봇 플랫폼을 제어할 수 있는 Vision-Language-Action 모델이며, Gemini Robotics-ER 1.5는 embodied reasoning에서 최첨단 성능을 달성하는 Vision-Language 모델이다.

Achievement

Figure 1 | The Gemini Robotics 1.5 family of models consists of Gemini Robotics 1.5, a VLA, and Gemini

• Multi-embodiment 제어: Motion Transfer를 통해 ALOHA, Bi-arm Franka, Apollo humanoid 로봇을 로봇 특화 post-training 없이 동일 checkpoint로 제어
• Zero-shot skill transfer: 한 로봇에서 학습한 기술을 명시적 재학습 없이 다른 로봇으로 전이
• Embodied thinking 능력: VLA가 행동 이전에 자연어 기반 추론 과정을 생성하여 복잡한 다단계 작업 분해 및 해석 가능성 향상
• 최첨단 embodied reasoning: Gemini Robotics-ER 1.5가 시각-공간 이해, 작업 계획, 진행률 추정 등 embodied reasoning 벤치마크에서 최고 성능 달성
• Agentic 시스템 통합: VLM orchestrator와 VLA action model의 결합으로 외부 도구 활용, 안전 메커니즘, 장기 작업 수행 가능

How

Figure 1 | The Gemini Robotics 1.5 family of models consists of Gemini Robotics 1.5, a VLA, and Gemini

• Motion Transfer (MT) 메커니즘: 다양한 로봇 소스로부터 학습하여 통일된 운동 및 물리적 상호작용 이해 형성
• Thinking VLA: 시각 관찰을 언어 기반 사고로 변환하고 context window에 thinking trace를 append하여 행동 생성
• Embodied Thinking: VLM과 VLA 모두에 적용되는 추론 능력으로 high-level planning과 low-level action decomposition 지원
• Agentic framework: GR-ER 1.5 (orchestrator)가 사용자 입력 처리, 작업 분해, 성공 감지를 수행하고 GR 1.5 (action model)를 tool로 호출
• Multi-embodiment dataset: ALOHA, Bi-arm Franka, Apollo 플랫폼의 다양한 조작 기술 데이터 수천 개로 학습