The advantage of such movements is that they are very fast. The lack of control therein, however, normally leads to the situation that behaving systems form a closed-loop with their environment where sensory inputs influence motor output, which in turn will create different sensations. Let's get back to our example of driving a car on a curvy road. In this example the view of the curve segment generates visual input to the system and steering is one possible output. Clearly, our perception of the road (steepness of the curve) influences how much we have to steer, whereas turning the steering wheel will cause changes in our perception for the next time moment. Visually guided reaching and grasping, navigation in the environment, servoing in robots are also examples of such closed-loop systems. Different from open-loop systems, closed-loop systems can react to unexpected situations and/or adapt to environmental changes by ways of learning.

We investigate closed-loop learning systems where the emphasis is on the development and utility of receptive fields in a closed-loop behavioural context. A receptive field (RF) of a given neuron is that particular surface area of a sensory organ from which neuronal responses can be elicited. Or in other words, the collection of sensors which form synapses to a single neuron form the neuron's receptive field. For example, the RF of a ganglion cell in the retina of the eye is composed of inputs from photoreceptors which provide its input, whereas a group of ganglion cells in turn forms the RF for a cell in the brain (Kandel et al., 2000). Receptive fields are found in different brain regions such as visual, somatosensory and auditory cortex. Another type of receptive fields are place fields (PFs) found in rat hippocampus (O'Keefe and Dostrovsky, 1971). Place fields of pyramidal cells code for a specific location of the animal in its environment. Like other receptive fields, PFs are formed from sensory inputs but diff.er from conventional RFs in that PFs are formed from multiple sensory cues such as visual, olfactory, somatosensory, auditory and self motion cues (Knierim et al., 1995; Save et al., 1998, 2000; Hill and Best, 1981; Etienne and Jeffery, 2004). The novelty of our approach is that we simultaneously develop and use receptive fields in behavioural tasks creating a closed-loop scenario. We form receptive fields from sensory inputs where at the same time RFs are used to drive the behaviour of the agent. When acting in the environment, sensory inputs change, which in turn influence the formation of the receptive fields closing the loop.