MoMaGen:

Generating Demonstrations under Soft and Hard Constraints for Multi-Step Bimanual Mobile Manipulation

Chengshu Li*, Mengdi Xu*, Arpit Bahety*, Hang Yin*, Yunfan Jiang, Huang Huang, Josiah Wong, Sujay Garlanka, Cem Gokmen, Ruohan Zhang, Weiyu Liu, Jiajun Wu, Roberto Martín-Martín, Fei-Fei Li
Stanford University, University of Texas at Austin
*Indicates Equal Contribution

arXiv Code Poster Video

Abstract

Imitation learning from large-scale, diverse human demonstrations has been shown to be effective for training robots, but collecting such data is costly and time-consuming. This challenge intensifies for multi-step bimanual mobile manipulation, where humans must teleoperate both the mobile base and two high-DoF arms. Prior X-Gen works have developed automated data generation frameworks for static (bimanual) manipulation tasks, augmenting a few human demos in simulation with novel scene configurations to synthesize large-scale datasets. However, prior works fall short for bimanual mobile manipulation tasks for two major reasons: 1) a mobile base introduces the problem of how to place the robot base to enable downstream manipulation (reachability) and 2) an active camera introduces the problem of how to position the camera to generate data for a visuomotor policy (visibility). To address these challenges, MoMaGen formulates data generation as a constrained optimization problem that satisfies hard constraints (e.g., reachability) while balancing soft constraints (e.g., visibility while navigation). This formulation generalizes across most existing automated data generation approaches and offers a principled foundation for developing future methods. We evaluate on four multi-step bimanual mobile manipulation tasks and find that MoMaGen enables the generation of much more diverse datasets than previous methods. As a result of the dataset diversity, we also show that the data generated by MoMaGen can be used to train successful imitation learning policies using a single source demo.

Video

Overview

Given a single source demonstration, as well as annotations for object-centric subtasks for each end-effector, MOMAGEN first randomizes scene configuration, and transforms the end-effector poses from the source demo to the new objects' frame of reference. For each subtask, it tries to sample a valid base pose that satisfies reachability and visibility constraints. Once found, it plans a base and torso trajectory to reach the desired base and head camera pose while trying to look at the target object during navigation. Once arrived, it plans an arm trajectory to the pregrasp pose and uses task space control for replay, before retracting back to a tucked, neutral pose

MoMaGen Overview

Data Generation

Data generation using MoMaGen for D2 randomization. D2 applies extremely aggressive scene randomization: task-relevant objects are placed freely on existing furniture without constraints, and additional obstacles are added. To our knowledge, no prior work has tackled data generation at this level of scene variability.

Pick Cup

Tidy Table

Clean Pan

Put Dishes Away

Policy Rollout (Simulation)

Pick Cup (D0)

Pick Cup (D1)

Tidy Table (D0)

Policy Rollout (Real-world)

WB-VIMA: Trained from scratch on 40 real demos


WB-VIMA: Pretrained on 1k MoMaGen data, finetuned on 40 real demos

Pi0: Finetuned on 40 real demos

Pi0: Finetuned on 1k MoMaGen sim data and 40 real demos

Cross-Embodiment Data Generation

MoMaGen enables cross-embodiment data generation. In this example, a single demonstration collected on a Galexea R1 robot is used to generate trajectories for a TIAGo robot. By planning and replaying dense end-effector trajectories instead of joint-space paths, the approach remains largely agnostic to robot-specific kinematics. These results showcase the robustness and flexibility of our framework across platforms.

Data Collection on R1 Robot

Data Generation on Tiago Robot

BibTeX


        @misc{li2025momagengen,
              title={MoMaGen: Generating Demonstrations under Soft and Hard Constraints for Multi-Step Bimanual Mobile Manipulation}, 
              author={Chengshu Li* and Mengdi Xu* and Arpit Bahety* and Hang Yin* and Yunfan Jiang and Huang Huang and Josiah Wong and Sujay Garlanka and Cem Gokmen and Ruohan Zhang and Weiyu Liu and Jiajun Wu and Roberto Martín-Martín and Li Fei-Fei},
              year={2025},
              eprint={2510.18316},
              archivePrefix={arXiv},
              primaryClass={cs.RO},
              url={https://arxiv.org/abs/2510.18316}, 
        }