Max Robotics — Global Service Robot Directory + US Certification

About

Qingkai Lu is a leading researcher in robotic manipulation, whose work centers on the intersection of deep learning, grasp planning, and 3D geometric reasoning. His major contributions lie in redefining how robots approach multi-fingered grasping—a notoriously complex problem requiring precise coordination of multiple contact points. Lu pioneered a paradigm shift by formulating grasp planning as probabilistic inference within learned deep neural networks, moving beyond traditional sampling-based methods. His landmark 2019 paper, "Planning Multi-fingered Grasps as Probabilistic Inference in a Learned Deep Network" (64 citations), and its 2020 extension (56 citations) introduced a voxel-based 3D convolutional neural network that predicts grasp success from both visual object data and grasp configurations. This differentiable framework enables efficient, gradient-based optimization, outperforming prior sampling approaches. Further advancing the field, Lu’s 2020 work on "Learning Continuous 3D Reconstructions for Geometrically Aware Grasping" (7 citations) explicitly incorporates full 3D geometry into grasp selection, enabling robots to reason about unseen object structure. With cumulative citations exceeding 145, Lu’s research has profoundly impacted robotic manipulation, offering a principled, data-driven pathway to dexterous, multi-fingered grasping that bridges perception and action.