Robotics · 2026-10-14

ForceWAM: Decoupling Future and Response Learning from Demo and Play Data for Contact-Rich Manipulation

Zijun Xu, Wendi Chen, Xinkai Wang, Jiachen Zou, Hongjin Chen, Hairuo Liu, Bokai Lin, Shihao Ma, Ke Ma, Zhiwei Zhang, Zhuohao Yan, Ke Wu, Haoming Song, Xueqi Li, Chuan Wen, Jieru Zhao, Wenchao Ding, Cewu Lu, Lixin Yang

arXiv:2610.18852 PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

ForceWAM is a force world action model for contact-rich manipulation that decouples learning task-compatible interaction outcomes from learning how local actions change contact. A force world model predicts future video and six-axis force/torque trajectories (visual-wrench futures) from task demonstrations. A separate inverse dynamics model uses current observations plus recorded or generated visual-wrench futures to predict robot actions, updated reactively with fresh wrench feedback. Real-robot results reach 85–95% mean success.

Why it matters

The authors address contact-rich manipulation where visual similarity can correspond to different force/contact states. By explicitly conditioning execution on predicted future visual-wrench trajectories and training inverse dynamics with both task demonstrations and task-agnostic play, ForceWAM aims to improve goal-directed contact regulation beyond what reactive baselines trained mainly on task data can do.

Method

ForceWM (force world model) jointly predicts future wrist video frames and six-axis wrench trajectories from successful task demonstrations using a pretrained video diffusion backbone.
IDM (inverse dynamics model) maps current observations and a visual-wrench future (recorded from task/play windows or generated at deployment) to action sequences, conditioned on action chunks and using within-chunk wrench feedback updates.
Deployment uses slow-fast execution: ForceWM refreshes predicted visual-wrench references periodically, while IDM streams action predictions and re-denoises using cached noise and growing measured wrench feedback between updates.

Abstract (from arXiv)

Expert demonstrations show how to complete contact-rich tasks, but provide limited coverage of how interaction forces respond to different actions. Task-agnostic play provides broader experience of contact dynamics without specifying task goals. To fully leverage these two types of data, we introduce ForceWAM, a force world action model that decouples knowledge of task-compatible interaction outcomes from knowledge of how local actions change contact. A world model learns from successful demonstrations to jointly predict future video and six-axis force/torque trajectories. Together, these futures describe the expected evolution of motion and contact during the task. A separate inverse dynamics model (IDM) learns from both task and play, mapping current observations and recorded visual-wrench futures to their associated robot actions. Play data thus provide knowledge of force responses for goal-directed execution. At deployment, the IDM reactively translates generated futures into actions, reusing them with fresh wrench feedback between world-model updates. Across real-robot force-regulated sliding, box flipping with and without disturbances, and ring placement, ForceWAM achieves $85$-$95\%$ mean success, surpassing world-modeling and reactive baselines. Our experiments further show that adding task-agnostic play data improves success by $15$-$45$ percentage points, demonstrating its value for learning contact-rich execution. Project page and code are available at https://force-wam.github.io/.

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC

preprint