Benezit, Florence ; Thiran, Patrick ; Vetterli, Martin

Presented at: ICASSP 2009, Taipei, Taiwan, April 19-24, 2009.

In: IEEE, ICASSP 2009, 2009, p. 3661 - 3664

Date: 2009

We design distributed and quantized average consensus algorithms on arbitrary connected networks. By construction, quantized algorithms cannot produce a real, analog average. Instead, our algorithm reaches consensus on the quantized interval that contains the average. We prove that this consensus in reached in finite time almost surely. As a byproduct of this convergence result, we show that the majority voting problem is solvable with only 2 bits of memory per agent.

Keyword(s): Consensus, quantization, gossip, voting

Reference: LCAV-CONF-2009-003

URL: http://www.icassp09.com/default.asp