Driving · 2026-10-01
UniWAM Technical Report: Unified Mobile Manipulation via Mixed-Stream World-Action Modeling and Manipulation Anchor Pose Supervision
Wei Xue, Keliang Liu, Mingzhang Cui, Jinhua Xie, Jinjie Wei, Jianan Hou, Jingcheng Lu, Lintao Wang, Kaixiang Qiu, Yizhou Liu, Xinghai Ye, Jinghang Han, Mingcheng Li, Jie Gu, Shunli Wang, Lihua Zhang, Dingkang Yang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
UniWAM is a unified mobile-manipulation world-action model that models navigation and manipulation as two independently sampled streams over a shared backbone. It uses separate action encoders and output heads, enabling either stream-only inference or both streams in one batch. The paper introduces Manipulation Anchor Pose (MAP) supervision and constructs MAP-Data (1.5M+ episodes, 7,500 hours) from large-scale 3D scenes to supervise where to stop and how to orient for manipulation.
Why it matters
The authors address two practical issues in unified mobile manipulation: (1) architectural coupling that wastes supervision on stationary commands by requiring complete trajectories for both behaviors in one sample, and (2) scarce data that explicitly supervises manipulation-ready terminal poses. MAP-Data provides dense image-plane targets (target boxes and MAP coordinates) and dense navigation supervision along approach trajectories, allowing the navigation stream to learn terminal stopping/orientation for manipulation explicitly and at scale.
Method
- UniWAM uses mixed-stream world-action modeling: separate navigation and manipulation streams with their own visual inputs, action encoders, and output heads, while sharing video DiT and an action DiT via mixture-of-transformers attention.
- Streams are sampled independently and packed as separate batch rows with attention that does not cross batch rows, so each sample supervises only its own stream and enables independent stream inference.
- Manipulation Anchor Pose (MAP) supervision defines an explicit terminal base position and heading for manipulation; MAP-Data is built from large-scale 3D scenes and provides per-frame projected MAP coordinates and target-object bounding boxes, plus manipulation targets via projected end-effector trajectories.
Limitation
The authors report that UniWAM’s navigation stream can predict small residual base motion during manipulation stage, moving the base slightly (suppressed by their near-zero rule).
Abstract (from arXiv)
Mobile manipulation requires precise navigation to a manipulation-ready pose followed by reliable object interaction. These two stages differ in action spaces and visual requirements, which complicates unified policy learning. In addition, collecting diverse real-world navigation data with explicit manipulation-ready pose supervision remains costly and difficult to scale. We introduce UniWAM, a unified mixed-stream world-action model with separate action encoders and output heads for navigation and manipulation, sharing a common backbone. This design supports joint representation learning on independently sampled navigation and manipulation data. UniWAM supports independent inference for either stream and batch-parallel inference for both. We further introduce Manipulation Anchor Pose (MAP) supervision for where to stop and how to orient for manipulation. An automated pipeline constructs MAP-Data from large-scale 3D scenes, yielding over 1.5 million episodes and 7,500 hours. MAP-Data provides per-frame target-object bounding boxes and image-plane MAP coordinates as auxiliary navigation supervision. Together with projected end-effector trajectories for manipulation, these prediction targets provide stream-specific image-plane supervision for action learning from egocentric observations. With large-scale MAP-Data, UniWAM outperforms the strongest external baselines on our MAP-Bench by 30.1\% in position error and 44.0\% in heading error. Across 24 real-robot tasks, UniWAM achieves leading results in MAP navigation and mobile manipulation, with competitive manipulation performance. We have released code, data, and benchmark.
Related papers
- UNITAS: A 3D-Native World Action Model for Embodied Manipulation
- PointWAM: 3D World Action Modeling for Dexterous Robotic Manipulation
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- UMI-WM: Interactive World Modeling for Bimanual Manipulation from Wrist Cameras
- SkeleWAM: Skeleton World-Action Modeling for Efficient Robotic Manipulation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-09-30 · Indexed by Pub-AI: 2026-10-01 UTC
UniWAM Technical Report: Unified Mobile Manipulation via Mixed-Stream World-Action Modeling and Manipulation Anchor Pose Supervision
Preprint posted: 2026-09-30 · Original record ↗
- arxiv
- 2609.39388
Explicit source versions (1)
- v1 · 2026-09-30
- Preprint available
Source date: 2026-09-30 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Mobile manipulation requires precise navigation to a manipulation-ready pose followed by reliable object interaction. These two stages differ in action spaces and visual requirements, which complicates unified policy learning. In addition, collecting diverse real-world navigation data with explicit manipulation-ready pose supervision remains costly and difficult to scale. We introduce UniWAM, a unified mixed-stream world-action model with separate action encoders and output heads for navigation and manipulation, sharing a common backbone. This design supports joint representation learning on independently sampled navigation and manipulation data. UniWAM supports independent inference for either stream and batch-parallel inference for both. We further introduce Manipulation Anchor Pose (MAP) supervision for where to stop and how to orient for manipulation. An automated pipeline constructs MAP-Data from large-scale 3D scenes, yielding over 1.5 million episodes and 7,500 hours. MAP-Data provides per-frame target-object bounding boxes and image-plane MAP coordinates as auxiliary navigation supervision. Together with projected end-effector trajectories for manipulation, these prediction targets provide stream-specific image-plane supervision for action learning from egocentric observations. With large-scale MAP-Data, UniWAM outperforms the strongest external baselines on our MAP-Bench by 30.1\% in position error and 44.0\% in heading error. Across 24 real-robot tasks, UniWAM achieves leading results in MAP navigation and mobile manipulation, with competitive manipulation performance. We have released code, data, and benchmark.
