Adaptive Collective Deliberation Models for Distributed Autonomous Systems
ID：157
View protection：Participant Only
Updated time：2026-07-27 13:16:14 Views：6
Online
Abstract
Abstract—Coordination of perception, planning and action among multiple agents for distributed autonomous systems increasingly depends on multi-agent governance mechanisms that are free of a single controlling agent. As the number of agents grows, existing methods – centralized coordinators, decentralized autonomous organization (DAO) style voting and federated multi-agent aggregation show difficulties in achieving such accuracy, trust, latency and resilience simultaneously. We propose a model of governance based on repeated exchange of proposals, bounded deliberation rounds, and dynamically calculated scores of trust and competence per agent, which are used for weighting in the process to achieve consensus, dubbed Adaptive Collective Deliberation Model (ACDM). ACDM is formalized with 7 equations that explain the evolution of trust, the adaptive weighting mechanism, the aggregation of beliefs, convergence, adaptive round control, energy cost, and composite governance efficiency index. A discrete event simulator was employed to compare ACDM with centralized and DAO-based governance and federated governance in terms of their performance for populations of 10, 20, ..., 200 agents. At N = 200, ACDM achieved up to 24.8% increase in the accuracy of the decision, decreased latency by an average of 63.6%, and increased the resilience score from 0.51 to 0.92. The results show that this adaptive trust-weighted deliberation can achieve a desirable level of accuracy, scalability and fault tolerance for decentralized autonomous systems.
Keywords
adaptive deliberation; distributed governance; multi-agent systems; trust modelling; decentralized decision-making; autonomous systems; consensus algorithms
Post comments