FPGA implementation of learning for online system identification

Artificial Neural Networks have been used in this paper to identify systems. Training online such networks using a gradient-based learning algorithm is one way to identify a system model. This task as a real-time embedded application requires the design of effective computational architectures which provide attractive trade-off between area and throughput. In this paper, we propose an FPGA implementation of an Artificial Neural Network with gradient-based learning for system identification. The design methodology is based on custom design of processing elements and the use of look-up tables for pre-computation of arithmetic operations. The results of implementation on a Virtex-5 XUPV5-LX110T evaluation platform are reported to demonstrate the efficacy of this approach.