Robotic hands tend to have a high number of sensors and actuators in a small space, whose supervising and control must be performed at the same time with precision and, in some cases, at high speed. Usually, a grid of Micro-Processors-Units or Micro-Controller-Units (MCUs) is employed to solve this problem, given the ease of programming. However, this solution can carry some drawbacks like the necessity of more space for computer resources. This work introduces a System-on-Chip based approach that carries out the control of a robotic hand with multiple Degree of Freedom. In order to justify its advantages, the proposed solution was tested in a robotic hand and compared against other implementations on (a) an ATMEL microcontroller, (b) an ARM processor, and (c) a full-dedicated hardware architecture on FPGA. Comparisons were made in terms of performance, computational resources, and power consumption. Additionally, our results showed an improvement over a previous MCU-Grid-based hand controller that achieved a control loop frequency of 1 KHz. In contrast, the proposed SoC approach achieved a 47.26 kHz frequency, showing the advantages of using our parameterizable floating-point arithmetic cores, which allow the designer to adjust the word-width to optimize both the hardware resources and energy consumption.

机械手往往在狭小的空间中具有大量的传感器和执行器，它们的监视和控制必须同时精确且在某些情况下以高速执行。通常，考虑到编程的简便性，采用微处理器单元或微控制器单元（MCU）的网格来解决此问题。但是，此解决方案可能会带来一些缺点，例如需要更多的计算机资源空间。这项工作介绍了一种基于片上系统的方法，该方法可以控制具有多个自由度的机械手。为了证明其优势，所提出的解决方案经过机械手测试，并与（a）ATMEL微控制器，（b）ARM处理器和（c）FPGA上的专用硬件体系结构的其他实现方案进行了比较。在性能，计算资源和功耗方面进行了比较。此外，我们的结果表明，与以前的基于MCU网格的手动控制器相比，该控制器实现了1 KHz的控制环路频率。相比之下，拟议的SoC方法实现了47.26 kHz的频率，显示了使用我们的可参数化浮点算术内核的优势，这使设计人员能够调整字宽以优化硬件资源和能耗。