Distributed MPC for Safe and Scalable Consensus of Heterogeneous Multi-Agent Systems in a Zero-Trust Environment

This paper proposes a trust-evaluation-based distributed model predictive control (DMPC) strategy for safe and scalable consensus of heterogeneous multi-agent systems (MASs) in a zero-trust environment. Dempster-Shafer theory (DST) is employed to quantify inter-agent trustworthiness and derive a global trust metric, thereby mitigating the impact of network threats and false testimonies during the trust evaluation process. Based on the computed trust, artificial reference trajectories are constructed to define the safe and scalable consensus state that each agent tracks, enabling adaptive regulation of reliance on neighbor information through real-time trust weights. The consensus problem is then reformulated as a trust-aware DMPC tracking problem that depends solely on locally received information, supporting distributed decision-making under zero-trust communication conditions. Sufficient conditions are derived to ensure recursive feasibility of the optimization problem, exponential stability of the closed-loop system, and the achievement of safe and scalable consensus. The effectiveness of the proposed strategy is validated via numerical simulations and vehicle platoon experiments.