Domain Randomization and Generative Models for Robotic Grasping

Tobin, Joshua W.D.; Biewald, Lukas; Duan, Rocky; Andrychowicz, Marcin; Handa, Ankur; Kumar, Vikas; Schneider, Jonas; Welinder, Peter; Zaremba, Wojciech; Abbeel, Pieter

doi:10.48550/arxiv.1710.06425

Cited by 12 publications

(15 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, CoinRun [4] and the broader Procgen Benchmark [5] use procedural generalization of environments at controllable levels of difficulty to demonstrate that effective generalization can require an extremely large number of training environments. Another manifestation of the generalization gap is the sim2real problem in robotics: agents trained in simulation overfit to this domain and fail to operate in hardware [6,7,8].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Invariant Policy Optimization: Towards Stronger Generalization in Reinforcement Learning

Sonar,

Pacelli,

Majumdar

2020

Preprint

View full text Add to dashboard Cite

A fundamental challenge in reinforcement learning is to learn policies that generalize beyond the operating domain experienced during training. In this paper, we approach this challenge through the following invariance principle: an agent must find a representation such that there exists an action-predictor built on top of this representation that is simultaneously optimal across all training domains. Intuitively, the resulting invariant policy enhances generalization by finding causes of successful actions. We propose a novel learning algorithm, Invariant Policy Optimization (IPO), that explicitly enforces this principle and learns an invariant policy during training. We compare our approach with standard policy gradient methods and demonstrate significant improvements in generalization performance on unseen domains for Linear Quadratic Regulator (LQR) problems and our own benchmark in the MiniGrid Gym environment.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Next, we relate the optimization problem (6) to a game played between n d players. Each player corresponds to a domain d and chooses a policy π d to maximize its own utility function R d (π av • Φ).…”

Section: Invariant Policy Optimizationmentioning

confidence: 99%

Invariant Policy Optimization: Towards Stronger Generalization in Reinforcement Learning

Sonar,

Pacelli,

Majumdar

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Alternatively, domain randomization has been proposed [25], where synthetic data is generated with sufficient variation at training time that the network is able to generalize to realworld data at test time. This process removes the need for domain adaptation, and has been used successfully for object detection and classification [13,25,26], and training robotic control processes [24,14]. This approach is particularly well suited to our application, since large, wellannotated datasets are not available, and randomization can be easily introduced in to the generation process of branching structures through structural changes, point jittering and dropout.…”

Section: Training On Synthetic Datamentioning

confidence: 99%

Deep Semantic Instance Segmentation of Tree-Like Structures Using Synthetic Data

Halupka

Garnavi

Moore

2019

2019 IEEE Winter Conference on Applications of Computer Vision (WACV)

View full text Add to dashboard Cite

Tree-like structures, such as blood vessels, often express complexity at very fine scales, requiring high-resolution grids to adequately describe their shape. Such sparse morphology can alternately be represented by locations of centreline points, but learning from this type of data with deep learning is challenging due to it being unordered, and permutation invariant. In this work, we propose a deep neural network that directly consumes unordered points along the centreline of a branching structure, to identify the topology of the represented structure in a single-shot. Key to our approach is the use of a novel multi-task loss function, enabling instance segmentation of arbitrarily complex branching structures. We train the network solely using synthetically generated data, utilizing domain randomization to facilitate the transfer to real 2D and 3D data. Results show that our network can reliably extract meaningful information about branch locations, bifurcations and endpoints, and sets a new benchmark for semantic instance segmentation in branching structures.

show abstract

“…Recently, domain randomization techniques are exploited to transfer action policies trained in simulation to the real world. These approaches randomize different aspects of the tasks, such as dynamics, [17], or visual sensory observations, [8], [18], [19], [20], We apply domain randomization to improve the versatility of the training samples collected from simple environments with few obstacles. The resulting dataset contains more complex configurations with multiple obstacles and challenging pathways.…”

Section: Domain Randomizationmentioning

confidence: 99%

Deep Reinforcement Learning to Acquire Navigation Skills for Wheel-Legged Robots in Complex Environments

Chen

Ghadirzadeh

Folkesson

et al. 2018

2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

Mobile robot navigation in complex and dynamic environments is a challenging but important problem. Reinforcement learning approaches fail to solve these tasks efficiently due to reward sparsities, temporal complexities and high-dimensionality of sensorimotor spaces which are inherent in such problems. We present a novel approach to train action policies to acquire navigation skills for wheel-legged robots using deep reinforcement learning. The policy maps heightmap image observations to motor commands to navigate to a target position while avoiding obstacles. We propose to acquire the multifaceted navigation skill by learning and exploiting a number of manageable navigation behaviors. We also introduce a domain randomization technique to improve the versatility of the training samples. We demonstrate experimentally a significant improvement in terms of data-efficiency, success rate, robustness against irrelevant sensory data, and also the quality of the maneuver skills.

show abstract

Domain Randomization and Generative Models for Robotic Grasping

Cited by 12 publications

References 0 publications

Invariant Policy Optimization: Towards Stronger Generalization in Reinforcement Learning

Invariant Policy Optimization: Towards Stronger Generalization in Reinforcement Learning

Deep Semantic Instance Segmentation of Tree-Like Structures Using Synthetic Data

Deep Reinforcement Learning to Acquire Navigation Skills for Wheel-Legged Robots in Complex Environments

Contact Info

Product

Resources

About