Human Limb Model Structure Optimization with Genetic Algorithm

Evolutionary approach is used in this research to adjust the structure of a human limb model and select the parameters related to the data acquisition. Portable device used to supervise therapeutic exercises imposes restrictions on the computational complexity allowed to model patient´s limbs which in turn narrows the choice of possible modeling techniques. While neural networks based models possess all properties necessary to model human limb dynamics their computational complexity may be too high. Genetic algorithm is used to optimize the structure of the neural networks based models of human limbs keeping delicate balance between the model quality and complexity of its structure. Structures of the the neural networks corresponding to the candidate limb models are encoded in the form of binary vectors then genetic algorithm is used to find most suitable structures.