We perform shape optimization of anti-resonance fiber (ARF) structures using a particle swarm optimization algorithm, achieving the lowest confinement loss of 2.4347×10−5 dB/m at 1.07 μm based on a single-layer tubing structure. Different from the existing design approaches that focus on size optimization and stacking of tubes, this research employs algorithm-driven point iteration and reconstruction to explore new structures for ARFs. This approach enhances the design flexibility and provides a new perspective for the design of ARFs. Furthermore, it was discovered that adding an additional tubing can effectively control the bending loss. This method enables the optimization of tubing shapes that are difficult to achieve through parameter analysis and supports local optimization of tubing structures. This method is capable of discovering higher-performance ARF structures based on existing classical designs, which is a significant inspiration for the design of non-uniform waveguides.