Analysis of minimal radial basis function network algorithm for real-time identification of nonlinear dynamic systems

A performance analysis is presented of the minimal resource allocating netw ork (MRAN) algorithm for online identification of nonlinear dynamic systems . Using nonlinear time-invariant and time-varying identification benchmark problems, MRAN's performance is compared with the online structural adaptiv e hybrid learning (ONSAHL) algorithm. Results indicate that the MRAN algori thm realises networks using fewer hidden neurons than the ONSAHL algorithm, with better approximation accuracy. Methods for improving the run-time per formance of MRAN for real-time identification of nonlinear systems are deve loped. An extension to MRAN is presented, which utilises a winner neuron st rategy and is referred to as the extended minimum resource allocating netwo rk (EMRAN). This modification reduces the computation load for MRAN and lea ds to considerable reduction in the identification time, with only a small increase in the approximation error. Using the same benchmark problems, res ults show that EMRAN is well suited for fast online identification of nonli near plants.