Fault-Resilient Learning-Assisted Predictive Control Protocols for Networked Crowd Navigation with Path Constraints
Year: 2026
Type: article
Abstract: Cyber-physical systems with networked crowd navigation necessitate distributed control protocols, which are safe, scalable, and resilient to uncertain communication among components and faults. Traditional centralised or entirely model-based methods are unable to ensure the satisfaction of the constraints and good stability of the coordination in thick network settings with fluid disturbances and inaccurate system representation. It is suggested in this paper that a fault-tolerant learning-based distributed model predictive control (DMPC) protocol against path, collision, and input constraints can be used to secure crowd navigation. The local agents solve a constrained predictive control problem based on the limited neighbour information shared over an unreliable communication network. In order to adjust model misfits and model disturbances, learning based residual dynamics estimator is incorporated in the DMPC prediction model to allow adaptive uncertainty reduction at the cost of distributed implement ability. The protocol-level based fault resilience mechanism identifies communication anomalies and actuator anomalies and reconfigures the local optimization targets and neighbour interaction to maintain the safety and feasibility. The suggested framework will ensure recursive feasibility, constraint satisfaction and closed-loop stability with limited disturbances and intermittent communication. Comprehensive simulations in dense crowd conditions indicate that navigation accuracy, collision avoidance and robustness in dense crowd scenarios has been vastly enhanced over baseline DMPC and centralised MPC schemes and most especially when fault rates and network degradation are very high. The findings show the usefulness of the developed protocol that ensures safe, scalable, and resilient crowd navigation in networked cyber-physical environments.
Author: Aakansha Soy
Institution: Kalinga University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7220163336