Inverse kinematics of a 6 DoF human upper limb using ANFIS and ANN for anticipatory actuation in ADL-based physical Neurorehabilitation

Abstract

Objective: This research is focused in the creation and validation of a solution to the inverse kinematics problem for a 6 degrees of freedom human upper limb. This system is intended to work within a realtime dysfunctional motion prediction system that allows anticipatory actuation in physical Neurorehabilitation under the assisted-as-needed paradigm. For this purpose, a multilayer perceptron-based and an ANFIS-based solution to the inverse kinematics problem are evaluated. Materials and methods: Both the multilayer perceptron-based and the ANFIS-based inverse kinematics methods have been trained with three-dimensional Cartesian positions corresponding to the end-effector of healthy human upper limbs that execute two different activities of the daily life: 'serving water from a jar' and 'picking up a bottle'. Validation of the proposed methodologies has been performed by a 10 fold cross-validation procedure. Results: Once trained, the systems are able to map 3D positions of the end-effector to the corresponding healthy biomechanical configurations. A high mean correlation coefficient and a low root mean squared error have been found for both the multilayer perceptron and ANFIS-based methods. Conclusions:The obtained results indicate that both systems effectively solve the inverse kinematics problem, but, due to its low computational load, crucial in real-time applications, along with its high performance, a multilayer perceptron-based solution, consisting in 3 input neurons, 1 hidden layer with 3 neurons and 6 output neurons has been considered the most appropriated for the target application.

Intelligent trajectory conversion and inverse dynamic control of a 3-DOF neuro-rehabilitation platform

“…The normalized firing strengths are calculated from the fixed circular nodes in layer 3 as given in (7) and then multiplied with (7) the linear parametric equations (formed in terms of the inputs) in the adaptive square nodes of layer 4, see (8). (8) where are the consequent parameters.…”

“…Furthermore, finding a closed form solution has been particularly difficult, and even if exist, the solution may not be unique [3][4][5][6] due to the many possible robot configurations for a given end-effector position and orientation. Although several conventional methods namely the algebraic, geometric, and numeric approaches have been adopted to solve the inverse kinematics problem [5], they are still faced with the above mentioned limitations, and the solutions proffered, as in the case of the numeric approach, may be impractical for real time application [7,8].…”

“…Intelligent methods such as the ANN and ANFIS have been applied in recent studies to solve the inverse kinematics problem of robotic devices [6][7][8][9][10]. They generally present a low computational load and do not suffer from singularity problems [9].…”

IK-FA, a New Heuristic Inverse Kinematics Solver Using Firefly Algorithm

“…For these methods, the IK problem is addressed as an optimization and the heuristics is used as computing alternatives. The second class includes nature inspired techniques and hybrid techniques with learning capacities such as neural networks and neuro-fuzzy systems (Rutkowski et al, 2012;Juang, 2000;Zaidi et al, 2012;Pérez-Rodríguez et al, 2012) where the IK solver is observed as system with a set of inputs and output(s), the system needs first to be trained using a set of targets and solutions generally obtained using the forward kinematic model, it is then validated with a separate validation set and finally used. When the target point is far from the training set this systems generate limited precision solutions compared to Jacobian based conventional methods (Chiaverini et al, 1994;Buss, 2004); they are also time consuming and can not satisfy real time constraints (Zaidi et al, 2013;Pérez-Rodríguez et al, 2012).…”

“…The training process has a direct impact on the quality of the obtained IK solver, for these intelligent IK solvers designing a good training set is essential. Neuro-fuzzy has the advantage to be interpretable when compared to neuralnetwork IK solutions; they are accurate but suffer from computing time, and could not be used in real time applications (Pérez-Rodríguez et al, 2012, Tolani et al, 2000.…”