A Meta-Cognitive Swarm Intelligence Framework for Resilient UAV Navigation in GPS-Denied and Cluttered Environments

Autonomous navigation of UAV swarms in perceptually-degraded environments, where GPS is unavailable and terrain is densely cluttered, presents a critical bottleneck for real-world deployment. Existing optimization-based planners lack the resilience to avoid catastrophic convergence to local optima under such uncertainty. Inspired by principles of computational meta-cognition, this paper introduces a novel swarm intelligence framework that enables a fleet of UAVs to autonomously sense, adapt, and recover from planning failures in real-time. At its core is the Self-Learning Slime Mould Algorithm (SLSMA), which integrates three meta-cognitive layers: a situation-aware search strategy that dynamically selects between exploration and exploitation based on perceived search stagnation; a collective memory mechanism that allows the swarm to learn from and avoid previously failed trajectories; and an adaptive recovery behavior that triggers global re-exploration upon entrapment. We formulate the multi-UAV trajectory problem as a resilient planning challenge, with a cost function that penalizes not only path length and collisions but also navigational uncertainty and proximity to failure states. Extensive simulations in synthetically complex 3D worlds and against the CEC 2017 benchmark suite demonstrate the framework's superior performance. The SLSMA does not merely optimize paths; it generates resilient trajectories, demonstrating a 99.5% mission success rate and significantly outperforming state-of-the-art metaheuristics in recovery speed and solution reliability. This work provides a foundational step towards truly autonomous swarms capable of persistent operation in denied and dynamic environments.