A Probabilistic Concept Web on a Humanoid Robot

Çelikkanat, Hande; Orhan, Guner; Kalkan, Sinan

doi:10.1109/tamd.2015.2418678

Cited by 17 publications

(24 citation statements)

References 70 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• Finally, motivated by our findings of the computational advantages in [3], we propose applying LDA on a concept web representation of the scene, instead of its raw features directly. We subsequently demonstrate how learning context from high-level concepts, instead of raw features, is easier and achieves higher performance.…”

Section: B This Studymentioning

confidence: 99%

“…The current article uses the concept web model that we developed in [3]. This previous work introduced a concept web model and showed why it is important and useful.…”

Section: B This Studymentioning

confidence: 99%

“…In our system, context is formalized over a set of concepts that are extracted from the scene, and represented in a densely-connected web structure, called the concept web [3]. Since this web is central to our model, before continuing with the exact formalization and use of context in the system, we briefly describe the extraction of relevant concepts from a scene, and the formation of the concept web.…”

Section: A Concept Web Using Markov Random Fieldmentioning

confidence: 99%

“…For this, we employ a widely-used topic model in computational linguistics, called Latent Dirichlet Allocation (LDA), and apply it to the active concepts in the scene. For modeling the concepts, we use a concept-web model that we developed using Markov Random Fields in our previous work [3]. Table Lamp Living Room "Chair" Chair features (c) Our proposal that links context with active concepts in the web Fig.…”

Section: Introductionmentioning

confidence: 99%

“…1: (a) Existing cognitive systems have concepts which have links to perceptual features and motor actions which were programmed by a designer or trained in context-free environments. (b) The concept web model that we developed in our previous study [3]: A densely connected concept web connecting perception, action and language; however, there is no notion of context in this model. (c) We propose a system that learns in context the links between concepts and sensorimotor primitives, based on the statistics of its interactions in real-life environments.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Learning Context on a Humanoid Robot using Incremental Latent Dirichlet Allocation

Çelikkanat

Orhan

Pugeault

et al. 2016

IEEE Trans. Cogn. Dev. Syst.

Self Cite

View full text Add to dashboard Cite

Abstract-In this article, we formalize and model context in terms of a set of concepts grounded in the sensorimotor interactions of a robot. The concepts are modeled as a web using Markov Random Field, inspired from the concept web hypothesis for representing concepts in humans. On this concept web, we treat context as a latent variable of Latent Dirichlet Allocation (LDA), which is a widely-used method in computational linguistics for modeling topics in texts. We extend the standard LDA method in order to make it incremental so that (i) it does not re-learn everything from scratch given new interactions (i.e., it is online) and (ii) it can discover and add a new context into its model when necessary. We demonstrate on the iCub platform that, partly owing to modeling context on top of the concept web, our approach is adaptive, online and robust: It is adaptive and online since it can learn and discover a new context from new interactions. It is robust since it is not affected by irrelevant stimuli and it can discover contexts after a few interactions only. Moreover, we show how to use the context learned in such a model for two important tasks: object recognition and planning.

show abstract

Section: B This Studymentioning

confidence: 99%

“…The current article uses the concept web model that we developed in [3]. This previous work introduced a concept web model and showed why it is important and useful.…”

Section: B This Studymentioning

confidence: 99%

Section: A Concept Web Using Markov Random Fieldmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Learning Context on a Humanoid Robot using Incremental Latent Dirichlet Allocation

Çelikkanat

Orhan

Pugeault

et al. 2016

IEEE Trans. Cogn. Dev. Syst.

Self Cite

View full text Add to dashboard Cite

show abstract

COSMO: Contextualized scene modeling with Boltzmann Machines

Bozcan

Kalkan

2019

Robotics and Autonomous Systems

Self Cite

View full text Add to dashboard Cite

Scene modeling is very crucial for robots that need to perceive, reason about and manipulate the objects in their environments. In this paper, we adapt and extend Boltzmann Machines (BMs) for contextualized scene modeling. Although there are many models on the subject, ours is the first to bring together objects, relations, and affordances in a highly-capable generative model. For this end, we introduce a hybrid version of BMs where relations and affordances are incorporated with shared, tri-way connections into the model. Moreover, we introduce a dataset for relation estimation and modeling studies. We evaluate our method in comparison with several baselines on object estimation, out-ofcontext object detection, relation estimation, and affordance estimation tasks.Moreover, to illustrate the generative capability of the model, we show several example scenes that the model is able to generate, and demonstrate the benefits of the model on a humanoid robot. The code and the dataset are publicly made available at: https://github.com/bozcani/COSMO

show abstract

Reciprocal Learning for Robot Peers

Liu

Kuo

et al. 2017

IEEE Access

View full text Add to dashboard Cite

This paper proposes a robot peer reciprocal learning system in which robot peers can not only cooperatively accomplish a difficult task but also help each other to learn better. In this system, each robot is an independent individual and has the ability to make individual decisions. They can communicate about image information, individual decisions, and current state to formulate mutual decisions and motions. For learning a new concept, we propose a mutual learning method, which allows the robots to learn from each other by exchanging weights in their neural network concept learning system. The simulation results show that the robots can learn from each other to build general concepts from limited training, while improving both of their performances at the same time. Finally, we design two cooperative tasks, which require the robots to formulate mutual sequential motions and keep communicating to manage their motions. The robotic experiments demonstrate that the proposed robot peer reciprocal learning system can help robots achieve difficult tasks in appropriate and cooperative ways, just as humans do.INDEX TERMS Mutual concept learning, reciprocal learning, robot cooperation, robot peer.

show abstract

A Probabilistic Concept Web on a Humanoid Robot

Cited by 17 publications

References 70 publications

Learning Context on a Humanoid Robot using Incremental Latent Dirichlet Allocation

Learning Context on a Humanoid Robot using Incremental Latent Dirichlet Allocation

COSMO: Contextualized scene modeling with Boltzmann Machines

Reciprocal Learning for Robot Peers

Contact Info

Product

Resources

About