A Magnetic Dipole Hybrid Localization Algorithm Based on Multilayer Perceptron and Genetic Algorithm

Non-line-of-sight localization of medical instruments with permanent magnets can effectively solve the medical dilemma of being unable to confirm the position during conventional detection visually. At present, magnetic tracking technology uses sensors to obtain the magnetic induction strength values and input them into the algorithm to calculate the orientation and posture of the permanent magnet. However, conventional localization algorithms inevitably have a high delay or low precision. This paper proposes a new permanent magnet positioning method - a hybrid algorithm based on multilayer perceptron and genetic algorithm. Multilayer perceptron provides an excellent initial solution for the genetic algorithm, which avoids the dilemma of the genetic algorithm falling into a locally optimal solution. Meanwhile, the genetic algorithm is compatible with the initial error of the neural network, which can quickly iterate in a small range. Theoretical and experimental results show that the proposed method can achieve localization and tracking within $8\text{cm}\times 8\text{cm}\times 8\text{cm}$ space at an average operating frequency of $\geq 10\text{Hz}$ , while the average position and orientation errors are controlled within 0.5mm and 5°.