Topology Aware Convention Emergence

Single convention convergence across different types of networks is a challenging multi-agent task. Our central hypothesis in this paper is that no simple distributed mechanism (such as the state-of-the-art Generalized Simple Majority (GSM) rule) can achieve this. We augment the agents with "network thinking" capability to solve this single convention convergence problem. Topological features such as node degree is used to design the accumulated coupling strength (ACS) convention selection algorithm. However, ACS does not perform as well in random networks as GSM does. Hence we propose a topology aware convention selection (TACS) algorithm that enables the agents to predict their local neighborhood topology and then to select a suitable convention selection algorithm. We have performed an extensive simulation study on random and SF networks showing that the majority of the agents correctly recognize their topology and use the appropriate convention mechanism leading to the convergence into a single convention.