While several Mobile Sink (MS)-based data gathering methods have been proposed for Wireless Sensor Networks, most of them are less adaptive to changes in network topology, and the planned MS trajectory cannot be refined to accommodate node failures. Hence, a KD-Tree-based scheme (KDT) is proposed, which is an adaptive and robust algorithm that reduces network energy consumption and data gathering delay. In contrast to many existing algorithms, the number of Rendezvous Points (RPs) are assigned dynamically by prioritizing the nodes’ coverage. Overlapping coverage of RPs is minimized, while guaranteeing 100% coverage of nodes. KDT is adaptive to network topology changes, and the planned MS trajectory can be refined to accommodate node failures. The shortest MS path is found using Ant Colony Optimization. Simulation results show that KDT requires approximately half the number of RPs and about 13% reduction in MS travel time compared to existing schemes.