ABSTRACT
Despite recent demonstrations that deep learning methods can successfully recognize and categorize objects using high dimensional visual input, other recent work has shown that these methods can fail when presented with novel input. However, a robot that is free to interact with objects should be able to reduce spurious differences between objects belonging to the same class through motion and thus reduce the likelihood of overfitting. Here we demonstrate a robot that achieves more robust categorization when it evolves to use proprioceptive sensors and is then trained to rely increasingly on vision, compared to a similar robot that is trained to categorize only with visual sensors. This work thus suggests that embodied methods may help scaffold the eventual achievement of robust visual classification.
- R. D. Beer. The dynamics of active categorical perception in an evolved model agent. Adaptive Behavior, 11(4):209--243, 2003.Google ScholarCross Ref
- Y. Bengio. Learning deep architectures for AI. Foundations and Trends® in Machine Learning, 2(1):1--127, 2009. Google ScholarDigital Library
- J. Bongard. The utility of evolving simulated robot morphology increases with task complexity for object manipulation. Artificial Life, 16(3):201--223, 2010. Google ScholarDigital Library
- J. Bongard. Morphological and environmental scaffolding synergize when evolving robot controllers: artificial life/robotics/evolvable hardware. In Proceedings of the 13th annual conference on Genetic and Evolutionary Computation, pages 179--186. ACM, 2011. Google ScholarDigital Library
- J. Bongard. Morphological change in machines accelerates the evolution of robust behavior. Proceedings of the National Academy of Sciences, 108(4):1234--1239, 2011.Google ScholarCross Ref
- J. C. Bongard. Evolutionary robotics. Communications of the ACM, 56(8):74--83, 2013. Google ScholarDigital Library
- E. Coumans et al. Bullet Physics Library, version 2.82. http://bulletphysics.org, 2014.Google Scholar
- M. Dorigo and M. Colombetti. Robot Shaping: Developing Autonomous Agents Through Learning. Artificial Intelligence, 71(2):321--370, 1994. Google ScholarDigital Library
- M. H. Fischer and R. A. Zwaan. Embodied language: A review of the role of the motor system in language comprehension. The Quarterly Journal of Experimental Psychology, 61(6):825--850, 2008.Google ScholarCross Ref
- P. Fitzpatrick, G. Metta, L. Natale, S. Rao, and G. Sandini. Learning about objects through action: initial steps towards artificial cognition. In IEEE Intl Conf on Robotics and Automation, volume 3, pages 3140--3145. IEEE, 2003.Google ScholarCross Ref
- N. Hansen. Cma evolution strategy, 1.1.05. https://www.lri.fr/ hansen/cmaes\_inmatlab.html, 2014.Google Scholar
- S. Harnad. To cognize is to categorize: Cognition is Categorization. Handbook of Categorization in Cognitive Science, pages 20--45, 2005.Google ScholarCross Ref
- G. E. Hinton. Learning multiple layers of representation. Trends in Cognitive Sciences, 11(10):428--434, 2007.Google ScholarCross Ref
- G. Lakoff and R. E. Núñez. Where mathematics comes from: How the embodied mind brings mathematics into being. Basic Books, 2000.Google Scholar
- M. Lungarella, G. Metta, R. Pfeifer, and G. Sandini. Developmental robotics: a survey. Connection Science, 15(4):151--190, 2003.Google ScholarCross Ref
- A. Nguyen, J. Yosinski, and J. Clune. Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. IEEE CVPR, 2015.Google ScholarCross Ref
- S. Perkins and G. Hayes. Robot shaping: Principles, methods and architectures. 1996.Google Scholar
- J. Plumert and P. Nichols-Whitehead. Parental scaffolding of young children. Developmental Psychology, 32(3):523--32, 1996.Google ScholarCross Ref
- L. M. Saksida, S. M. Raymond, and D. S. Touretzky. Shaping robot behavior using principles from instrumental conditioning. Robotics and Autonomous Systems, 22(3):231--249, 1997.Google ScholarCross Ref
- C. Scheier and R. Pfeifer. Classification as sensory-motor coordination. In Advances in Artificial Life, pages 657--667. Springer, 1995. Google ScholarDigital Library
- C. Szegedy, W. Zaremba, I. Sutskever, J. Bruna, D. Erhan, I. Goodfellow, and R. Fergus. Intriguing properties of neural networks. International Conference on Learning Representations, 2014.Google Scholar
- E. Tuci, G. Massera, and S. Nolfi. Active categorical perception of object shapes in a simulated anthropomorphic robotic arm. IEEE Transactions on Evolutionary Computation, 14(6):885--899, 2010. Google ScholarDigital Library
Index Terms
- An Embodied Approach for Evolving Robust Visual Classifiers
Recommendations
Evolving autonomous specialization in congested path formation task of robotic swarms
Redundancy in the number of robots is a fundamental feature of robotic swarms to confer robustness, flexibility, and scalability. However, robots tend to interfere with each other in a case, where multiple robots gather in a spatially limited ...
Evolving collective step-climbing behavior in multi-legged robotic swarm
AbstractThis paper focuses on generating the collective step-climbing behavior of a multi-legged robotic swarm. Most studies on swarm robotics develop collective behaviors in a flat environment using mobile robots equipped with wheels. However, these ...
Evolving Self-Organizing Behaviors for a Swarm-Bot
In this paper, we introduce a self-assembling and self-organizing artifact, called a swarm-bot, composed of a swarm of s-bots, mobile robots with the ability to connect to and to disconnect from each other. We discuss the challenges involved in ...
Comments