A new approach to robotics is described in my paper published in the Artificial Life journal, June 2017.
The approach is based on the concept that the behaviour of living systems is a process of controlling perceptions, rather than controlling or computing (behavioural) output. In principle, the complex behaviour that we observe emerges just from a hierarchy of these simple feedback control systems, and accounts for all types and levels of behaviour.
Applying this perspective to robotics has significant implications, such as explicit, predictive models are not required, the systems are inherently adaptive and are not prone to unpredictable disturbances or changes in system parameters (e.g. weight or length of limbs), they are computationally lightweight and are comprised of a simple, universal process.