Robotics: Science and Systems XV 2019
DOI: 10.15607/rss.2019.xv.077
|View full text |Cite
|
Sign up to set email alerts
|

Abstract: Algorithms based on deep network models are being used for many pattern recognition and decision-making tasks in robotics and AI. Training these models requires a large labeled dataset and considerable computational resources, which are not readily available in many domains. Also, it is difficult to understand the internal representations and reasoning mechanisms of these models. The architecture described in this paper attempts to address these limitations by drawing inspiration from research in cognitive sys… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
34
0

Year Published

2019
2019
2022
2022

Publication Types

Select...
5
2
1

Relationship

8
0

Authors

Journals

citations
Cited by 19 publications
(34 citation statements)
references
References 25 publications
0
34
0
Order By: Relevance
“…However, the underlying capability of modeling and reasoning about intentional actions is relevant to other problems and applications characterized by dynamic changes. For instance, other work within our research group has combined the reasoning capabilities of our architecture with inductive learning of domain constraints to guide the construction of deep networks that have been used for estimating the occlusion and stability of object structures [19] and for answering explanatory questions about images [21,22]; other research groups have explored the combination of ASP-based knowledge representation with low-level perceptual processing for explaining spatial relations in videos [28]. Future research can adapt our architecture to such problems in more complex domains to demonstrate the scalability and wider applicability of our architecture.…”
Section: Discussion and Future Workmentioning
confidence: 99%
“…However, the underlying capability of modeling and reasoning about intentional actions is relevant to other problems and applications characterized by dynamic changes. For instance, other work within our research group has combined the reasoning capabilities of our architecture with inductive learning of domain constraints to guide the construction of deep networks that have been used for estimating the occlusion and stability of object structures [19] and for answering explanatory questions about images [21,22]; other research groups have explored the combination of ASP-based knowledge representation with low-level perceptual processing for explaining spatial relations in videos [28]. Future research can adapt our architecture to such problems in more complex domains to demonstrate the scalability and wider applicability of our architecture.…”
Section: Discussion and Future Workmentioning
confidence: 99%
“…The labeled examples are also used to train decision trees and incrementally learn previously unknown state constraints that are used for subsequent reasoning. Experimental results show a marked improvement in accuracy and computational efficiency in comparison with an architecture that only uses deep networks, and provides insights about the interplay between reasoning and learning; for complete details, see [Mota and Sridharan, 2019a].…”
Section: Motivationmentioning
confidence: 99%
“…Figure 1(left) shows our architecture in the context of a Robot Assistant (RA) domain, with a simulated robot estimating the occlusion of objects and the stability of object structures, and rearranging objects to reduce clutter. Spatial relations between objects in RGB-D images are grounded using our prior work [Mota and Sridharan, 2018]. An object is occluded if any fraction of its frontal face is hidden by another object; a structure is unstable if any object in it is unstable.…”
Section: Proposed Architecturementioning
confidence: 99%
“…This architecture builds on our prior work on combining commonsense inference with deep learning (Riley and Sridharan, 2018a;Mota and Sridharan, 2019) by introducing the ability to learn and reason with constraints governing domain states, and extending explainable inference with commonsense knowledge to also support planning and diagnostics to achieve any given goal. Although we use VQA as a motivating example, it is not the main focus of our work.…”
Section: Introductionmentioning
confidence: 99%