FPGA implementation of Pseudorandom Number Generators with a generic 2-D cellular automata architecture

There are computational models such as Monte Carlo simulation, game-based computing theory, neural networks and genetic algorithms, in which pseudorandom numbers play an important role. These numbers must have some good statistical properties such as normal distribution, uncorrelated and a long period. In this paper a Pseudorandom Numbers Generator (PRNG) based on 2-dimensional cellular automata with asymmetric neighborhood is proposed. Our approach is based a configurable architecture that offers the ability to parameterize the size of the grid, transition rules, number of neighbors, seed and boundary conditions. The architecture is targeted to Field Programmable Gate Array (FPGA) technology. Several models were analyzed in terms of the quality of randomness, as well as performance and resource consumption. Results show that the generated sequences have passed standard statistical tests, making the architecture suitable for embedded applications.