A wooden tank measuring 1.0 m �� 1.0 m �� 0.70 m was used. After the pipe was positioned, its supports were removed, and it was then covered with dry soil. The surface of the tank was covered with a polypropylene plate. Eleven paths parallel to the x-axis and another eleven paths parallel to the y-axis were marked on this plate. These 22 paths were
Human beings, and their ancestors before them, have evolved throughout millions of years and obviously their systems to perform tasks too. Most of these tasks are commanded by the brain. Therefore, engineers, and specially the neuromorphic engineering community [1,2] have fixed as their main goal to mimic the human systems which are supposed to have an extraordinary behavior carrying out their own tasks.

In particular, reaching movements (planning and execution) have been for ages one of the most important and studied ones [3]. If we take a closer look in humans, we will find that the system involved in these tasks is the central nervous system (CNS). This system is a combination of the brain and the spinal cord and, simplifying, it consists of neuron cells and uses spikes or graduated potentials to transmit on the information across the anatomy [3].Nowadays, it is possible to integrate several thousands of artificial neurons into the same electronic device (very-large-scale integration (VLSI) chip [4], Field-Programmable Gate Array (FPGA) [5] or Field-Programmable Analog Array (FPAA) [6]); which are called neuromorphic devices.

There are many European projects focused on building computing systems which exploit the capabilities GSK-3 of these devices (Brain-inspired multiscale computation in neuromorphic hybrid systems (BrainScale; website: http://brainscales.kip.uni-heidelberg.de/index.html), SpiNNaker (website: http://apt.cs.man.ac.uk/projects/SpiNNaker/) and the Human Brain Project (HBP; website: https://www.humanbrainproject.eu/) as examples). One of the main challenges is which devices and how to integrate them to produce functional elements.One of the problems faced when we try to integrate and implement these neural architectures is the communication between them: it is not easy to distinguish which neuron of what device is firing a spike. To solve this problem, new communication strategies have been exploited, such as the Address-Event- Representation (AER) protocol [7].

AER maps each neuron with a fixed address which is transmitted through the interconnected neuronal architecture. By using the AER protocol, all neurons of a layer are continuously sharing their excitation with the other layers through bus connections; this information can be processed in real time by a higher layer.AER was proposed to achieve communication between neuromorphic devices. It tries to mimic the structure and information coding of the brain.