TossingBot: Learning to Throw Arbitrary Objects With Residual Physics

By Andy Zeng, Shuran Song, Johnny Lee, Alberto Rodriguez, Thomas A. Funkhouser. 2019

460 citations recorded by the source corpus.

We investigate whether a robot arm can learn to pick and throw arbitrary rigid objects into selected boxes quickly and accurately. Throwing has the potential to increase the physical reachability and picking speed of a robot arm. However, precisely throwing arbitrary objects in unstructured settings presents many challenges: from acquiring objects in grasps suitable for reliable throwing, to handling varying object-centric properties (e.g., mass distribution, friction, shape) and complex aerodynamics. In this work, we propose an end-to-end formulation that jointly learns to infer control parameters for grasping and throwing motion primitives from visual observations (RGB-D images of arbitrary objects in a bin) through trial and error. Within this formulation, we investigate the synergies between grasping and throwing (i.e., learning grasps that enable more accurate throws) and between simulation and deep learning (i.e., using deep networks to predict residuals on top of control parameters predicted by a physics simulator). The resulting system, TossingBot, is able to grasp and successfully throw arbitrary objects into boxes located outside its maximum reach range at 500+ mean picks per hour (600+ grasps per hour with 85% throwing accuracy); and generalizes to new objects and target locations.

Subfields: Biomedical Engineering, Control and Systems Engineering, AI for Physical Sciences, Robotics & Control

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biomedical Engineering

In Control and Systems Engineering

In AI for Physical Sciences

In Robotics & Control

Citations in the corpus

MOReL : Model-Based Offline Reinforcement Learning (2020, methodology): “MBRL enables the use of generic priors like smoothness and physics [22] for model learning, and a wide variety of planning algorithms [23, 24, 25, 26, 27].”
RoboNet: Large-Scale Multi-Robot Learning (2019, background): “Deep neural network models have been used widely in a range of robotics applications [12, 13, 14, 15, 16, 17].”
Implicit Behavioral Cloning (2021, background): “, compounding errors [2, 3]), in practice it enables some of the most compelling results of real robots generalizing complex behaviors to new unstructured scenarios [4, 5, 6].”
SAPIEN: A SimulAted Part-Based Interactive ENvironment (2020, background): “PyBullet [10], a real-time physics engine with Python interface, powers a series of projects focusing on robotics tasks [60, 27].”
A comprehensive review of digital twin — part 1: modeling and twinning enabling technologies (2022, background): “The initial predictions can also be made directly by a physics-based model (Zeng et al., 2020).”
A Distributional View on Multi-Objective Policy Optimization (2020, background): “, 2016), perform complex robotic manipulation tasks (Zeng et al., 2019), learn agile locomotion (Tan et al.”
Clear Grasp: 3D Shape Estimation of Transparent Objects for Manipulation (2019, methodology): “We adapted a state-of-the-art grasping algorithm for our experiment [60, 59, 61, 58, 55], which consists of a convolutional network that predicts the probability of picking success for a scripted grasping primitive across a dense pixelwise sampling of end effector locations and orientations across t”
Catch & Carry (2019, methodology): “…is to design controllers that incorporate domain knowledge for some tasks and to then use learning to refine the behavior around this initial, engineered policy – this scheme has been applied to efforts in robotics for tossing objects [Zeng et al. 2019] and catching objects [Kim et al. 2014].”
Learning to Walk in the Real World with Minimal Human Effort (2020, background): “grasping tasks, by collecting a large amount of interaction data on real robots [37, 33, 57].”
Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulation (2022, background): “Also, instead of just predicting poses, the agent could also be trained to predict other physical parameters like target velocities [109].”
Learning to Rearrange Deformable Cables, Fabrics, and Bags with Goal-Conditioned Transporter Networks (2020, unknown): “This technique [67] has shown promising results for robotic warehouse order picking [71], pushing and grasping [70], tossing objects [69], kit assembly [66], and mobile manipulation [62].”
“Good Robot!”: Efficient Reinforcement Learning for Multi-Step Visual Tasks with Sim to Real Transfer (2019, methodology): “DRL has proven effective at increasingly complex tasks in robotic manipulation [1], [5], [22], [23].”
Keypoints into the Future: Self-Supervised Correspondence in Model-Based Reinforcement Learning (2020, background): “Another approach, taken by [29, 30, 23], is to rotate the observation into a frame deﬁned by the action.”
Learning Visual Affordances for Robotic Manipulation (2019, unknown): “This work appeared previously as Zeng et al. (2019) [126].”
Meta-Adaptive Nonlinear Control: Theory and Algorithms (2021, background): “Examples abound in robotics, such as a drone flying in different wind conditions [1], a manipulator throwing varying objects [2], or a quadruped walking over changing terrains [3].”
ReLMoGen: Leveraging Motion Generation in Reinforcement Learning for Mobile Manipulation (2020, methodology): “RL has been successfully applied to solve visio-motor tasks dealing with continuous control based on high dimensional observations [5, 6, 7, 8, 9, 10, 11].”