SA-Net: Robust State-Action Recognition for Learning from Observations

Soans, Nihal; Asali, Ehsan; Hong, Yi; Doshi, Prashant

doi:10.1109/icra40945.2020.9197393

Cited by 26 publications

(17 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep learning also opens the possibility of tackling the learning problem in an end-to-end manner. A recent step in this direction is SA-Net (Soans et al 2020), which is able to recognize state-action pairs using an RGB-D sensor. However, it still needs to be coupled with a policy learning algorithm.…”

Section: Mapping Sensors To Actuatorsmentioning

confidence: 99%

A Bayesian tracker for synthesizing mobile robot behaviour from demonstration

Magnenat

Colas

2021

Auton Robot

View full text Add to dashboard Cite

Section: Mapping Sensors To Actuatorsmentioning

confidence: 99%

A Bayesian tracker for synthesizing mobile robot behaviour from demonstration

Magnenat

Colas

2021

Auton Robot

View full text Add to dashboard Cite

“…To overcome the obstacle of noisy labels or features, perturbation schemes and data augmentations have been investigated. In computer vision, data augmentation is done by applying operations like cropping and rotation to combat potential mislabelled training data [2,17,12]. Another line of work achieves robustness against noisy data by generating data synthesizers that achieves the same predictive performance as using the real data.…”

Section: Data Augmentation and Synthetic Data Generation For Ro-mentioning

confidence: 99%

CrowdTeacher: Robust Co-teaching with Noisy Answers & Sample-specific Perturbations for Tabular Data

Sotoodeh,

Xiong,

2021

Preprint

View full text Add to dashboard Cite

Samples with ground truth labels may not always be available in numerous domains. While learning from crowdsourcing labels has been explored, existing models can still fail in the presence of sparse, unreliable, or differing annotations. Co-teaching methods have shown promising improvements for computer vision problems with noisy labels by employing two classifiers trained on each others' confident samples in each batch. Inspired by the idea of separating confident and uncertain samples during the training process, we extend it for the crowdsourcing problem. Our model, CrowdTeacher, uses the idea that perturbation in the input space model can improve the robustness of the classifier for noisy labels. Treating crowdsourcing annotations as a source of noisy labeling, we perturb samples based on the certainty from the aggregated annotations. The perturbed samples are fed to a Co-teaching algorithm tuned to also accommodate smaller tabular data. We showcase the boost in predictive power attained using CrowdTeacher for both synthetic and real datasets across various label density settings. Our experiments reveal that our proposed approach beats baselines modeling individual annotations and then combining them, methods simultaneously learning a classifier and inferring truth labels, and the Co-teaching algorithm with aggregated labels through common truth inference methods.

show abstract

“…Various methods exist to tackle the RL optimization problem. Among them, Policy Gradient techniques have enabled to use RL in real-world robotic contexts (see [31] for an extensive definition of these approaches and [32,33,34] for successful real-world robotic applications). These techniques are based on the use of a stochastic policy formally denoted by π θ , where θ is a vector of parameters.…”

Section: Model-free Adaptive Controlmentioning

confidence: 99%

Learning-based vs Model-free Adaptive Control of a MAV under Wind Gust

Chaffre¹,

Moras²,

Chan-Hon-Tong³

et al. 2021

Preprint

View full text Add to dashboard Cite

Navigation problems under unknown varying conditions are among the most important and well-studied problems in the control field. Classic model-based adaptive control methods can be applied only when a convenient model of the plant or environment is provided. Recent model-free adaptive control methods aim at removing this dependency by learning the physical characteristics of the plant and/or process directly from sensor feedback. Although there have been prior attempts at improving these techniques, it remains an open question as to whether it is possible to cope with real-world uncertainties in a control system that is fully based on either paradigm. We propose a conceptually simple learning-based approach composed of a full state feedback controller, tuned robustly by a deep reinforcement learning framework based on the Soft Actor-Critic algorithm. We compare it, in realistic simulations, to a model-free controller that uses the same deep reinforcement learning framework for the control of a micro aerial vehicle under wind gust. The results indicate the great potential of learning-based adaptive control methods in modern dynamical systems.

show abstract

SA-Net: Robust State-Action Recognition for Learning from Observations

Cited by 26 publications

References 17 publications

A Bayesian tracker for synthesizing mobile robot behaviour from demonstration

A Bayesian tracker for synthesizing mobile robot behaviour from demonstration

CrowdTeacher: Robust Co-teaching with Noisy Answers & Sample-specific Perturbations for Tabular Data

Learning-based vs Model-free Adaptive Control of a MAV under Wind Gust

Contact Info

Product

Resources

About