Numerical P Systems (NPS) is a variant of P systems using real-valued quantities. It was shown that it can successfully be applied for different real-world problems, in particular in the area of robotic control. In this paper we introduce an extension of NPS, called Generalized Numerical P Systems (GNPS) and we describe an efficient implementation of GNPS using FPGA hardware. This allows to build fast controller chips based on (G)NPS and interacting directly with the environment. Two test cases are presented describing the implementation results of Sobel image edge detection algorithm