Addressing Uncertainties in Decentralized Context Models of Autonomous Robot Teams

Autonomous robot teams operating in dynamic, uncertain environments require reliable mechanisms to build decentralized context models without centralized coordination. Traditional consensus methods often fail under uncertainty caused by inconsistent sensing, communication delays, or heterogeneous perception models. This paper introduces the Decentralized Belief Consensus (DBC) algorithm, a novel approach that integrates probabilistic reasoning with entropy-based certainty measures to enable adaptive and robust consensus formation in heterogeneous multi-robot systems. Each robot quantifies the uncertainty of its local observations using Shannon entropy, derives a certainty score, and fuses beliefs with neighbors through certainty-weighted averaging. This allows the team of autonomous robots to defer commitment when evidence is weak and dynamically adjust influence according to observation reliability. The DBC algorithm was evaluated through various simulations involving heterogeneous teams of unmanned aerial vehicles (UAV) and umanned ground vehicles (UGV) tasked with mine detection under varying levels of noise, false detections, and team sizes. Results demonstrate that DBC achieves high accuracy, full consensus rates, and strong robustness while maintaining competitive convergence times compared to established algorithms such as LCP, WMSR, CDCI, DBBS, and EEV. By explicitly modeling uncertainty in both sensing and communication, DBC provides a scalable foundation for reliable decentralized context modeling and collective perception in autonomous robot teams.