Robot policies trained on visually narrow demonstrations can become sensitive to environmental features that are unrelated to the task. Existing generative augmentation methods can introduce visual diversity, but processing every frame and camera view independently is computationally expensive and may produce inconsistent observations across cameras and timestamps. We introduce Augmentron, a geometry-aware pipeline that generates reusable scene representations and renders them through recorded camera poses. This produces temporally and multi-view consistent environ- ment and workspace variations while preserving the robot, task objects, actions, and annotations. Across three physical manipulation tasks, replacing half of the real world training dataset with Augmentron data improves model success across three tasks for both in out-of-distribution distractor objects, and out-of-distribution workspace variation evaluations. In a separate comparison against alternate augmentation techniques, Augmentron achieves the highest pooled success in all three conditions. The results show that introducing augmented data to a data training recipe can help policies generalize to dynamic real-world settings.
Figures & tables
Figure 1: Environmental and workspace surface visual diversity produced by Augmentron: a geometry-aware pipeline that generates multi-view and temporally consistent data augmentation scalably for robotics manipulation datasets.
Figure 2: Overview of Augmentron . Task annotations and robot geometry identify regions that must remain unchanged. A generated scene representation is then rendered and composited with the source demonstration to produce environment or workspace augmentations.
Figure 3: The five transformations used by the masked-noise baseline. Each corruption is applied outside the protected robot and task-object regions.
ID
OOD distractors
OOD workspace
Steps
Recipe
Effective volume
Cup n=50
Towel n=50
Vase n=50
Pooled n=150
Cup n=50
Towel n=50
Vase n=50
Pooled n=150
Cup n=50
Towel n=50
Vase n=50
Pooled n=150
30k
Real
1×
86
68
46
66.7
16
26
12
18.0
78
54
20
50.7
Aug. 50/50
1×
72
88
70
76.7
52
52
32
45.3
80
94
28
67.3
Aug. 33/67
3×
80
80
44
68.0
76
64
12
50.7
62
88
26
58.7
45k
Aug. 33/67
3×
82
72
60
71.3
74
80
44
66.0
70
84
28
60.7
RoboEngine
2×
70
82
20
57.3
60
86
14
53.3
78
40
6
41.3
Table 1: Task and pooled success rates (%). Bold marks the highest observed task success rate within each condition and training-step group. Each task cell contains 50 rollouts and each pooled cell contains 150.
Figure 4: Pooled physical-rollout success. Left: 30k results; only Real versus Augmentron 50/50 is matched in data volume and training schedule. Right: 45k recipe comparison; Augmentron uses 3× effective volume and each baseline uses 2× . Each bar aggregates n=150 rollouts.
Vision-Language-Action (VLA) policies often rely strongly on the camera viewpoints seen during training, causing substantial performance degradation when deployed from unseen perspectives. Collecting demonstrations from sufficiently diverse physical viewpoints is expensive and still provides only sparse coverage of the viewpoint space. We introduce InfiNoVA, a data-augmentation framework that converts synchronized multi-camera demonstrations into a dense distribution of geometrically consistent training views. InfiNoVA reconstructs each manipulation trajectory as a time-varying 3D Gaussian representation and renders novel observations from sampled camera poses while preserving the original state-action correspondence. This explicit scene representation improves frame-level fidelity and temporal consistency while reducing task-critical hallucinations observed in generative novel-view synthesis. Across four real-world manipulation tasks, policies trained with InfiNoVA achieve 5.4x higher average success under unseen randomized viewpoints than both VISTA-based augmentation and the unaugmented policy. InfiNoVA further achieves 1.7x higher success than training directly on all five physical camera views. These results show that dense, geometrically grounded viewpoint augmentation provides a practical route toward camera-robust robot policies without modifying the underlying policy architecture.
Sai Puneeth Reddy Gottam, Elmar Rueckert, Vedant Dave
Chair of Cyber Physical Systems, Montanuniversität Leoben, Austria
Visuomotor policies for manipulation have demonstrated remarkable potential in modeling complex robotic behaviors, yet minor alterations in the robot's initial configuration and unseen obstacles easily lead to out-of-distribution observations. Without extensive data collection effort, these result in catastrophic execution failures. In this work, we introduce an effective data augmentation framework that generates visually realistic fisheye image sequences and corresponding physically feasible action trajectories from real-world eye-in-hand demonstrations, captured with a portable parallel gripper with a single fisheye camera. We introduce a novel Gaussian Splatting formulation, adapted to wide FoV fisheye cameras, to reconstruct and edit the 3D scene with unseen objects. We utilize trajectory optimization to generate smooth, collision-free, view-rendering-friendly action trajectories and render visual observations from corresponding novel views. Comprehensive experiments in simulation and the real world show that our augmentation framework improves the success rate for various manipulation tasks in both the same scene and the augmented scene with obstacles requiring collision avoidance.
Chuer Pan, Litian Liang, Dominik Bauer +4
Stanford University · Columbia University · Toyota Research Institute
Vision-language-action (VLA) policies have shown strong potential for general-purpose manipulation, yet they often fail on novel, out-of-distribution objects whose appearance or geometry deviates from the training distribution. The standard remedy is to collect multi-view teleoperation data for every failure case, but this scales poorly in both cost and time. We introduce Pose6DAug, a failure-driven data augmentation framework that turns a policy's own successful episodes into targeted demonstrations for its failure modes, without any new data collection. Our key insight is that each successful episode already encodes a physically valid action trajectory together with calibrated multi-view observations. By swapping only the manipulated object while preserving this trajectory, we obtain new and physically grounded demonstrations. However, naive 2D video editing breaks multi-view consistency and physical plausibility, particularly under heavy occlusion and egocentric viewpoints. Our method instead operates directly in 3D, anchoring the target object with an explicit mesh driven by a temporally coherent 6D pose trajectory, ensuring geometrically consistent renderings across all camera views. Fine-tuning a VLA on data augmented by our method improves success rates by 16.5% relative to the state-of-the-art baseline on novel objects, while preserving in-distribution performance. These results show that multi-view and physically consistent augmentation is a practical path to scalable VLA generalization.