GPLAC: Generalizing Vision-Based Robotic Skills Using Weakly Labeled Images

By Avi Singh, Larry Yang, Sergey Levine. 2017

23 citations recorded by the source corpus.

We tackle the problem of learning robotic sensorimotor control policies that can generalize to visually diverse and unseen environments. Achieving broad generalization typically requires large datasets, which are difficult to obtain for task-specific interactive processes such as reinforcement learning or learning from demonstration. However, much of the visual diversity in the world can be captured through passively collected datasets of images or videos. In our method, which we refer to as GPLAC (Generalized Policy Learning with Attentional Classifier), we use both interaction data and weakly labeled image data to augment the generalization capacity of sensorimotor policies. Our method combines multitask learning on action selection and an auxiliary binary classification objective, together with a convolutional neural network architecture that uses an attentional mechanism to avoid distractors. We show that pairing interaction data from just a single environment with a diverse dataset of weakly labeled data results in greatly improved generalization to unseen environments, and show that this generalization depends on both the auxiliary objective and the attentional architecture that we propose. We demonstrate our results in both simulation and on a real robotic manipulator, and demonstrate substantial improvement over standard convolutional architectures and domain adaptation methods.

Subfields: Artificial Intelligence, Control and Systems Engineering

How researchers see this paper

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

In Artificial Intelligence

In Control and Systems Engineering

Citations in the corpus

Self-Supervised Correspondence in Visuomotor Policy Learning (2019, background): “The model is close to [1]–[3] with the key architectural traits of having a few convolutional layers followed by a channel-wise spatial expectation operation, which has been widely used [10], [11], [25], [29], [36]–[38].”
Online Meta-Learning (2019, positive): “After the convolution layers, we use a spatial soft-argmax to extract learned feature points, an architecture that has previously been shown to be effective for spatial tasks (Levine et al., 2016; Singh et al., 2017).”
How to Close Sim-Real Gap? Transfer with Segmentation! (2020, background): “Several issues may contribute to the generation of this gap, including an inaccurate robot model or model parameter setting in simulation, hysteresis, joint frictions, delays in the transmission of the control signals, and noisy measurements of the state in the real robot Tobin et al. [2017]. Fig.”
W-TALC: Weakly-supervised Temporal Activity Localization and Classification (2018, background): “In computer vision, researchers have utilized weak labels to learn models for several tasks including semantic segmentation [18,28,63], visual tracking [69], reconstruction [52,25], video summarization [37], learning robotic manipulations [46], video captioning [41], object boundaries [29], place re”