The state space of any coupled brain-body-environment system is impossibly large, and no doubt contains an almost infinite number of aperiodic, unpredictable attractors ( Cvitanovic et al., 2016).

In fact, this could be the very problem evolution was trying to solve by allowing cognitively sophisticated organisms to emerge in the first place ( Barrett, 2018). However, as the field of robotics has no doubt learned the hard way, it becomes a challenging problem to solve the coordination problem between the hundreds to thousands of sensors and motors on such a body. Advancements made in the space of embodied agents would be hugely fruitful to the study of both artificial intelligence and cognitive science, as any behaviors successfully revealed will be inherently adaptive to real-world conditions and thus will not suffer to the limited application to “toy worlds” or “test settings.”Īccording to the REC view, complex cognitive behaviors would require a body with sufficient complexity to support them.

We believe there is a pressing need to expand the scope of these models, which so far have been restricted to demonstrations of “minimally cognitive behavior” ( Brooks, 1991 Barandiaran and Moreno, 2006 van Duijn et al., 2006). Here the mind is not isolated from the world, but distributed across brain, body, and environment perception of the world is direct and does not need to be inferred or synthesized through symbolic representations. One of these approaches is radical embodied cognition (REC), which treats the brain, body, and environment in terms of coupled dynamical systems and ecological psychology ( Chemero, 2011). Neurorobotic models are perfect candidates for proving the validity of embodied/dynamical approaches to cognition ( Hoffmann and Pfeifer, 2018).

Download chaos general