Affiliations:
Department of Information Technology, Faculty of Computing and Information Technology, Northern Border University, Rafha, Saudi Arabia
Creating scalable, secure, and real-time coordination in decentralized multi-agent systems remains a major challenge. Centralized planning approaches often suffer from limited scalability and weak robustness, while purely decentralized methods may lead to inefficient behavior or deadlocks. This paper proposes a hierarchical framework that integrates Decentralized Model Predictive Control (MPC) with an intention-sharing consensus protocol. The coordination problem is formulated as a constrained stochastic optimal control problem, and Sequential Convex Programming (SCP) is applied to efficiently solve the non-convex trajectory optimization problems encountered by individual agents at the local level. At the higher level, a consensus protocol enables agents to actively resolve conflicts and align their intentions, thereby reducing short-sighted decision-making. Extensive simulations on challenging benchmark scenarios show that the proposed approach achieves performance close to centralized methods and significantly outperforms existing reactive and predictive baselines in terms of success rate, efficiency, and safety. Furthermore, the method demonstrates strong robustness to communication delays, sensor noise, and model uncertainty, indicating its suitability for real-world applications.
Multi-agent systems, Decentralized model predictive control, Consensus protocols, Sequential convex programming, Trajectory optimization
https://doi.org/10.21833/ijaas.2026.03.020
Alrslani, F. A. F. (2026). Consensus-based model predictive control for scalable multi-robot coordination. International Journal of Advanced and Applied Sciences, 13(3), 198–206. https://doi.org/10.21833/ijaas.2026.03.020