CS Seminar: Embodied Evolution and Social Learning

Advertisements for Seminars taking place on campus.
Forum rules
This forum is now unmoderated.

CS Seminar: Embodied Evolution and Social Learning

Postby J. Borg » Mon Oct 24, 2016 12:24 pm

WEDNESDAY 26TH OCTOBER, 3:00PM, CR08 (Turing Lab, Colin Reeves Building)

Tea, Coffee and Biscuits available before, during and after the seminar

"Embodied Evolution, Selection Pressure and Social Learning"
Dr. Evert Haasdijk, Department of Computer Science - VU Amsterdam

Ficici et al. (1999) coined the phrase embodied evolution for evolutionary processes that are distributed over the robots in the population to allow them to adapt autonomously and continuously. Embodied evolution offers a unique opportunity for autonomous on-line adaptivity in robot collectives. The vision behind embodied evolution is one of collectives of truly autonomous robots that can adapt their behaviour to suit varying tasks and circumstances. Autonomy occurs at two levels: not only do the robots perform their tasks without external control, they also assess and adapt –through evolution– their behaviour without referral to external oversight and so learn autonomously.
Embodied evolution implies that robots must adapt to their environment as well as learn to perform some user-defined tasks(s). In the first part of my presentation, I will discuss recent advances made researching the interaction of environmental and task-defined selection pressures in embodied evolution with the MONEE system (Haasdijk et al., 2014). The second part of my presentation will consider social learning. That social learning gives rise to a Darwinian process has been famously argued by Dawkins when he introduced the idea of memes 1976. I will argue that social learning is in fact equivalent to embodied evolutionary systems and outline our research on social learning in the DREAM project. This project develops a cognitive architecture that incorporates dream-like processes to allow robots to re-describe their knowledge and experiences at more abstract levels. Thus, the robots can generalise their experiences to learn and adapt more efficiently. In a collective set ting, the robots can combine forces to learn jointly, i.e., to implement social learning. Social learning in the context of different levels of abstraction in the knowledge base poses an intriguing question: under which circumstances is it better for the robots to exchange knowledge at the abstract or at the immediate level?

Dawkins, R. (1976). The Selfish Gene. Oxford University Press, Oxford.
Ficici, S. G., Watson, R. A., and Pollack, J. B. (1999). Embodied Evolution: A Response to Challenges in Evolutionary Robotics. In Wyatt, J. L. and Demiris, J., editors, Proceedings of the Eighth European Workshop on Learning Robots, pages 14–22.
Haasdijk, E., Bredeche, N., and Eiben, a. E. (2014). Combining environment-driven adaptation and task-driven optimisation in evolutionary robotics. PloS one, 9(6):e98466.
J. Borg
Posts: 30
Joined: Fri Oct 23, 2009 3:58 pm

Return to Seminars

Who is online

Users browsing this forum: Google [Bot] and 5 guests