Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces GraspMeanFlow, an SE(3)-equivariant MeanFlow framework that enables efficient few-step generation of 6-DoF grasp poses by learning average velocities over finite time intervals. The method retains equivariance under few-step sampling, significantly improving the efficiency of grasp generation by achieving the same performance as iterative SE(3) flow models with only a single function evaluation. Experimental results demonstrate that GraspMeanFlow can enhance grasp success rates by up to 24.3 points in the few-step regime while maintaining the desired transformation properties with respect to object movements.
Achieving real-time 6-DoF grasp generation with a single function evaluation, GraspMeanFlow outperforms traditional methods that require multiple iterations.
Recent data-driven methods for synthesizing 6-DoF grasp poses use generative models to learn complex grasp pose distributions and generate diverse candidate poses. In particular, SE(3)-equivariant flow-based models generate grasp poses that transform consistently with object rotations and translations. However, these methods sample by iterative numerical integration, requiring tens of function evaluations per grasp and limiting their use in real-time manipulation. We propose GraspMeanFlow, an SE(3)-equivariant MeanFlow framework for few-step 6-DoF grasp generation. Our method learns the average velocity over a finite time interval, defined through the time-ordered exponential so that it reproduces exactly the rigid-body displacement accumulated over that interval. We prove that a point-cloud-conditioned distribution transported by an equivariant average-velocity flow map remains invariant, so equivariance is retained under few-step sampling, and we condition the field on a pair of times by lifting both to equivariant vectors, leaving the backbone otherwise unchanged. For stable training, we pair a flow-matching boundary term with either of two consistency terms: the differential MeanFlow identity, whose target requires a Jacobian-vector product, or an equivalent semigroup loss that avoids it. Experiments on ACRONYM show that a single function evaluation of GraspMeanFlow reaches the EMD that an iterative SE(3) flow model needs five steps to approach, that a second instantiation of the same framework improves grasp success by up to 24.3 points in the few-step regime, and that both generate grasp distributions transforming exactly with the object.