New interaction in autonomous decentralized control method for future social systems

Social systems will become much larger and more complex as ever more things are connected to networks in the future. In such scalable future social systems, the supplied resources must be utilized so as to maximize the total profit for the huge number of users of the resources. However, a centralized control method has difficulty calculating the maximum value of the total profit because of the high computational burden of the calculation and high cost for aggregating a huge amount of data from the users for the calculation. In this paper, we present an interaction in an autonomous decentralized control (ADC) method for maximizing total profit for the users while avoiding excessive consumption of shared resources that causes the resource price to soar. In the ADC using the proposed interaction, the computational burden is distributed, and each user communicates with only its neighboring users to exchange information. The user sends a request for a behavior, based on a result of a comparison with values of the information sent from its neighboring users, to only its neighboring users. A user autonomously decides its own behavior by following a majority of the requests. The ADC using the proposed interaction was evaluated in terms of the total monetary profit using our developed multi-agent-based simulator that includes 1024 factory-agent models. The evaluation results indicate that the ADC with the proposed interaction achieved 98.0% of a theoretical total monetary profit calculated by a centralized control method while each factory agent communicates with only its neighboring agents. In addition, the total monetary profits in ADC with the proposed interaction ranged from 95.8 to 98.3% of those theoretical total monetary profits even when the power charges greatly fluctuated. Moreover, a prototype of ADC using the proposed interaction was developed and tested. In this test, changes of the total monetary profit of this developed prototype were measured using the fluctuation of the power charges. The results of the evaluations and the measurements revealed that the ADC using the proposed interaction is suitable for scalable future social systems.