Neural Dynamics Approach to Cognitive Robotics - A Hands-on Summer School

The embodiment stance postulates that cognition cannot be understood as long as the links between cognitive processes and the sensory and motor surfaces are left unexamined. An understanding of cognition must also reflect that cognition takes place in organisms or agents who are situated in structured environments to which their bodies and nervous systems are specifically adapted.

Creating robotic demonstrations of cognitive process models is a powerful way to establish that principles of embodiment are addressed. In fact, the links to sensory and motor surfaces and immersion in the environment are difficult to probe without real-world robotic implementations.

Event dates: 9 – 14 September 2013

In robotics, the behavior-based approach has similarly emphasized direct links to simple sensory and motor systems. For roboticists, the challenge has consisted of scaling such systems up toward true cognition, including finding ways of how to enrich such systems with representations. Neuronal dynamics provides a a powerful theoretical language that may address both goals. It is based on neural principles, but also enables robotic implementation through simple interfaces with sensory and motor systems.

An obstacle to the wide-spread use of neuronal dynamics as a theoretical framework by both researchers in embodied cognition and in autonomous robotics is the seeming mathematical sophistication required to make use of these concepts. This school provides a hands-on and down-to-earth introduction to neuronal dynamics ideas and enables participants to become productive within this framework.

Structure of school:
Mornings consist of tutorial lectures that systematically introduce the relevant concepts and methods, including all required math.

A variety of robots are available, with e-pucks and the Khepera II robot vehicles being the main work-horses for projects. The programming environment is Matlab as well as Phython in some projects.

Elementary neural dynamics; attractor dynamics; link to neuronal networks; self-excitation and competition; Mathematical foundations of the Dynamic Field Theory (DFT); Elements of embodied cognition: detection, estimation, selection, working memory, change detection, sequence generation; Implementation issues: interfacing to sensors and motors, numerical integration; Attractor dynamics in robotics: obstacle avoidance, target acquisition; DFT in robotics: object representations, scene representations, spatial language, behavioral organization, reinforcement learning.

For further information please visit: