In Wireless Sensor Networks (WSNs), multiple mobile sinks reduce energy-consumption hotspots and extend network lifetime, but balancing energy across sinks is challenging under uneven node distributions and diverse trajectories. This paper proposes a framework that jointly optimises sink trajectories and workloads. Density-Based Spatial Clustering of Applications with Noise (DBSCAN) partitions the field, iterative k-means selects rendezvous points (RPs), and Ant Colony Optimisation (ACO) computes energy-efficient closed tours through each sink’s RPs. Across 50–300 nodes over 20 randomised scenarios each, the proposed density-aware partitioning reduces average sink path length by 22.7% relative to Fuzzy C-Means (FCM) significant at every size (p<10⁻⁶), and retains higher residual sink energy (up to 1.997 J of a 2.0 J budget). The results reveal a densitydependent lifetime crossover: DBSCAN extends lifetime by up to 45.6% over FCM for sparse deployments (N ≤100), whereas FCM wins for dense deployments (N ≥150); both substantially outperform fixed hexagonal-grid partitioning. These findings establish density-aware clustering as the preferred strategy for path- and energy-efficient collection in small-to-medium mobilesink WSNs, and characterise the regime in which it is most effective.