我的新闻

Monte Carlo Diffusion for Generalizable Learning-Based RANSAC

Jiale Wang, Chen Zhao, Wei Ke, Tong Zhang

Abstract：

Random Sample Consensus (RANSAC) is a fundamental approach for robustly estimating parametric models from noisy data. Existing learning-based RANSAC methods utilize deep learning to enhance the robustness of RANSAC against outliers. However, these approaches are trained and tested on the data generated by the same algorithms, leading to limited generalization to out-of-distribution data during inference. Therefore, in this paper, we introduce a novel diffusion-based paradigm that progressively injects noise into ground-truth data, simulating the noisy conditions for training learning-based RANSAC. To enhance data diversity, we incorporate Monte Carlo sampling into the diffusion paradigm, approximating diverse data distributions by introducing different types of randomness at multiple stages. We evaluate our approach in the context of feature matching through comprehensive experiments on the ScanNet and MegaDepth datasets. The experimental results demonstrate that our Monte Carlo diffusion mechanism significantly improves the generalization ability of learning-based RANSAC. We also develop extensive ablation studies that highlight the effectiveness of key components in our framework.

Efficient Equivariant Flow Policy with Acceleration Regularization

JianleiChang, Ruofeng Mei, Wei Ke, Xiangyu Xu

Abstract:

Generative modeling has emerged as a powerful approach for visuomotor policy learning, with diffusion models achieving strong results in robotic manipulation. However, they suffer from two major limitations: poor data efficiency and slow sampling due to iterative inference. While recent advances introduce equivariant architectures to address the former, slow sampling speed remains a challenge. We propose Efficient Equivariant Flow Policy (EEFlow), a generative policy learning framework based on flow matching, which models a continuous path from noise to action using ordinary differential equations (ODEs). We theoretically show that under an isotropic Gaussian prior and an equivariant velocity field, EEFlow preserves equivariance in the learned action distribution, promoting better generalization across symmetric states and reducing data requirements. To improve sampling efficiency, we introduce a second-order regularizer that penalizes acceleration. Since computing acceleration requires intractable marginal trajectories, we propose a novel surrogate loss that enables stable training using only readily available conditional trajectories. Evaluated on extensive manipulation tasks, EEFlow matches or exceeds the performance of baselines while offering fast inference, highlighting its potential for high-performance, efficient robotic control.