QAにおけるGaussian on-offエンコーディングの構造解析に対する適用性に関する検討

抄録

This study applies an efficient random encoding method whose uncertainty are controlled through Gaussian distribution to two-dimensional truss structure analysis based on quantum annealing (QA). As the basic operation model of QA, Quadratic Unconstrained Binary Optimization (QUBO) model adopts binary qubits as variables, which means that generally an additional encoding method should be adopted to extend their expression range to any real numbers. The encoding method applied in this study selects a certain number of digits within the range of real numbers, and then controls whether they work in the total summation of real number expression through the 0/1 quantum state. A Gaussian distribution are adopted in it to control the selection of digits to represent real numbers, ensuring its stability. Other three commonly used encoding methods (binary, one-hot and on-off encoding) are employed as comparison under different truss structures and load conditions. The results show that the Gaussian on-off encoding outperforms present encoding methods in structure analysis in terms of accuracy and stability.