Ant Colony Optimization (ACO) is a highly effective metaheuristic inspired by the foraging behavior of real ants. However, its performance significantly degrades in dynamic environments due to “pheromone stagnation,” where accumulated pheromone trails prevent the swarm from responding rapidly to environmental changes. We introduce ADNS-ACO (Ant Dendritic Neural System-based ACO), a novel framework that integrates a Spiking Neural Network (SNN) to modulate the exploration-exploitation balance through simulated antennal biosignals. Our approach utilizes a Leaky Integrate-and-Fire (LIF) model to process a biologically inspired approximation of local tactile intensity. When the environment undergoes a perturbation, the SNN triggers a neuromodulatory response via a non-linear gating mechanism that dynamically resets pheromone influence, enabling rapid re-optimization. Experiments conducted on high-complexity Dynamic Traveling Salesman Problem (DTSP) benchmarks with 60 cities reveal that ADNS-ACO achieves a 21.10% improvement in solution quality and a 77.3% reduction in recovery latency compared to classical rank-based ACO. Statistical validation across 30 independent trials confirms a highly significant improvement () and a large effect size (Cohen’s ). These results position ADNS-ACO as a superior, event-driven solution for real-time dynamic optimization, with promising implications for autonomous routing and swarm robotics.
Keywords: Ant Colony Optimization, Bio-inspired Algorithms, Dynamic Optimization, Neuromorphic Computing, Spiking Neural Networks
