Encoding of quantisation partitions optimised for distributed estimation

Abstract

A novel algorithm of designing independently operating quantisers at each node that optimise the system-wide metric (i.e. estimation error) for distributed estimation systems is presented. Assuming that each node can collect only its local measurement, the quantisation partitions constructed based on minimisation of the estimation error are not independently encodable at each node. To ensure a perfect independent encoding, a new concept of assigning multiple codewords to each partition is introduced and a simple iterative algorithm that generates a set of such codewords for each partition is proposed and demonstrated through experiments to achieve a significant performance improvement over typical designs and distributed novel designs recently published. Furthermore, the codeword set optimisation rules are suggested to reduce the cardinality of the sets for practical applications.