Inspired by killer whale hunting strategies, this study presents a biomimetic algorithm for controlled subgroup fission in swarms. The swarm agents adopt the classic social force model with some practical modifications. The proposed algorithm consists of three phases: cluster selection phase via a constrained K-means algorithm, driven phase with strategic agent movement, including center pushing, coordinated oscillation, and flank pushing by specialized driven agents, and judgment phase confirming subgroup separation using the Kruskal algorithm. Simulation results confirm the algorithm's high success rate and efficiency in subgroup division, demonstrating its potential for advancing swarm-based technologies.