Ruihai Wu

Postdoc Scholar at UC Berkeley. Doctor of Philosophy at Peking University.

Email: ruihai [at] berkeley.edu

Email  /  Google Scholar  /  Github

We propose LeHome, a framework that is able to simulate 6 categories of deformable objects in household scenarios with low-cost robots.

For the generalization and stable coordination of complex bimanual manipulation tasks, we propose to use correspondence from foundation models for efficient adaptation of affordance and action.

WorldComposer is a novel approach for generative high-fidelity simulation with digital cousins for generalizable robot learning and evaluation.

RoboWM-Bench is a manipulation-centric benchmark for embodiment-grounded evaluation of video world models. It converts generated behaviors from both human-hand and robotic manipulation videos into embodied action sequences and validates them through robotic execution.

When lifting 2D features to geometry-profound 3D space, challenges arise, such as long runtimes, multi-view inconsistencies, and low spatial resolution with insufficient geometric information. Therefore, we propose Part-Aware 3D Feature Field (PA3FF), a novel dense 3D feature with part awareness for generalizable articulated object manipulation.

Furniture assembly requires precise dual-arm coordination where one arm manipulates parts while the other provides collaborative support and stabilization. We propose A3D, a framework which adaptively identifies optimal support and stabilization locations on furniture parts.

We propose to learn point-level affordance to model the complex space and multi-modal manipulation candidates of garment piles, with novel designs for the awareness of garment geometry, structure, inter-object relations, and further adaptation.

We exploit the geometric generalization of point-level affordance, learning affordance aware of bimanual collaboration in geometric assembly with long-horizon action sequences.

We propose to learn dense visual correspondence for diverse garment manipulation tasks with category-level generalization using only one- or few-shot human demonstrations.

We explore the task of manipulating articulated objects within environment constraints and formulate the task of environment-aware affordance learning for manipulating 3D articulated objects, incorporating object-centric per-point priors and environment constraints.

We study geometric shape assembly by leveraging SE(3) Equivariance, which disentangles poses and shapes of fractured parts.

We study deformable object manipulation using dense visual affordance, with generalization towards diverse states, and propose a novel kind of foresightful dense affordance, which avoids local optima by estimating states’ values for longterm manipulation.

We study dense geometry-aware, interaction-aware, and task-aware visual action affordance and trajectory proposals for manipulating articulated objects.

We propose localize-assemble-predicate network (LAP-Net), decomposing visual relation detection (VRD) into three sub-tasks to tackle the long-tailed data distribution problem.