We support the spike processing method for all the algorithms used; that is the main claim of this paper: design, development and implementation of a spike-based processing control architecture and to avoid using an external computer for processing, with extremely low power consumption and AER communication.First of all we have to determine the elements to integrate according with the biological principles: image sensor, a hardware architecture where the CNS behavior is emulated and a robot to execute the movements. There are many sorts of problems to deal with in this selection: sensor must be a spiking retina, the architecture has to keep as many CNS features as possible, within only addition, subtraction and injection of spikes, and finally, the robotic platform will be made of motors which mimic the muscles.

The first element of the architecture is the image sensor. We have chosen a silicon retina, the dynamic vision sensor developed by the Tobi Delbruck research group [23]. It is a VLSI chip made of 128 �� 128 analog pixel firing spikes (with AER protocol) when a threshold is reached.Applying several processing layers to these events flow [24], a single event, which meets the center of an object, is isolated. Therefore, this event plays the role of the target position for the system, so the retina will deliver the reaching position to the architecture.The main part of the system turns around the VITE algorithm [11]. It was selected because it is inspired by the biological movement and was designed to mimic it.

It has been translated into the spikes domain using spike-based building blocks which add, subtract or inject spikes like the human neural system. This algorithm generates a non-planned trajectory and it needs a second algorithm to produce and control the Carfilzomib forces applied to the motors which mimic the muscles. In this paper we are focused on the first algorithm and therefore, no feedback is performed. Our aim is to evaluate the viability of translating the VITE completely into the spikes domain and applying it to a real robotic platform in order to enable the second algorithm and to close the control loop.In order to achieve the described goal, we have transformed the algorithm using existing spike processing blocks developed for our research group [21,25,26] and put them into MATLAB Simulink to test and adjust the blocks.

Afterwards, two FPGA based boards were used to allocate the blocks and mimic the biological structure (one for the brain and the other one for the spinal cord). The final archite
A micromechanical silicon resonant accelerometer converts the acceleration signals to be tested into the frequency variation of the resonator. Thus, the output is a quasi-digital signal. Moreover, the micromechanical silicon resonant accelerometer has the advantages of a wide dynamic range, strong anti-interference capacity and high stability.