Data Compression/Decompression of a Table Look-Up Controller Using a Network Structure

摘要

Preliminary results concerning the compression and decompression of a table-based controller for a continuous time system are presented. The proposed method considers the design of a neural network structure which classifies all elements in the control law table. Note that for a sufficiently large control law table, this network requires less microprocessor storage memory than the original control law talbe (compression) and can determine the desired control effort in real time (decompression). The weights of the network are determined by direct assignment rather than by training.