Experimental Gaming Analysis of ADMM Dynamic Distributed Optimization Algorithm

Application of efficient distributed optimization algorithms in multi-agent systems designed for socio-economic problems demands additional conditions for algorithms to be satisfied. In particular, the algorithm should be incentive compatible, i.e., the model of behaviour of any agent should be in concordance with interests of a particular economic subject represented by this agent in system. On an example of the resource allocation problem we study incentive compatibility of the famous algorithm for dynamic distributed optimization, the alternating direction method of multipliers (ADMM).