Exploiting the SoC FPGA Capabilities in the Control Architecture of a Quadruped Robot

The use of FPGAs in the field of robotics has the potential to create interesting results considering the former's assets and the latter's necessities. Motion control is highly demanding both in resources and data, while the number of sensors employed in robotic devices can be large. On the other hand, FPGAs have gained a renowned place in the field of digital circuit implementation media. Simultaneously, the vast number of peripherals they are equipped with, renders them as ideal for the management of multiple external signals. In this paper, a highly affordable control architecture has been implemented for the quadruped robot Laelaps II based on a SoC FPGA, in order to examine the use of FPGAs in demanding robotic applications. The proposed system has realtime characteristics and is capable of controlling more than eight actuators at very high frequencies, outperforming most motion control platforms existing in the market, while keeping the total cost low. To achieve this, signals from all the robots peripherals are handled in parallel by the FPGA at a very high frequency, leaving the embedded ARM processor free to allocate all its resources to motion control and user interfacing tasks. The whole hardware/ software architecture is analyzed and evaluated compared with state-of-the-art approaches, while a trotting experiment with the Laelaps II motion system is presented for validation purposes.