Quantized consensus for agents on digraphs

The available investigations about quantized average consensus typically assume agents be confined to evolve on balanced digraphs via quantized information exchange, thus the corresponding update matrices are doubly stochastic, which is very restrictive and brings about feasibility problem in practical applications. By dropping the doubly stochastic constraint for the update matrices, this paper studies the consensus seeking for a group of agents on general strongly connected digraphs, where agents' states are communicated (may be unidirectional) through logarithmic quantizer. Under mild assumptions on network topology, we derive an upper bound for the quantization precision to guarantee the weighted average preservation of the whole network.