Manipulation

We learn reward functions in unsupervised object keypoint space, to allow us to follow third-person demonstrations with model-based RL.

We design and demonstrate a new tactile sensor for in-hand tactile manipulation in a robotic hand.

We demonstrate visual control within 20 seconds on a robot with unknown morphology, from a single uncalibrated RGBD camera.

High-resolution tactile sensing together with visual approaches to prediction and planning with deep neural networks enables high-precision tactile servoing tasks.

We propose a low-cost compact easily replicable hardware stack for manipulation tasks, that can be assembled within a few hours. We also provide implementations of robot learning algorithms for grasping (supervised learning) and reaching (reinforcement learning). Contributions invited!

We propose a low-cost compact easily replicable hardware stack for manipulation tasks, that can be assembled within a few hours. We also provide implementations of robot learning algorithms for grasping (supervised learning) and reaching (reinforcement learning). Contributions invited!

By exploiting high precision tactile sensing with deep learning, robots can effectively iteratively adjust their grasp configurations to boost grasping performance from 65% to 94%.