Trajectory Tracking using a Bio-inspired Neural Network for a Low-Cost Robotic Articulator

Brain circuits in the cerebellum are considered as central processing units of movement control and coordination. Using an internal model controller, it is possible to reconstruct brain like structures that can predict trajectories or reaching arm tasks. In this study, we have developed a bio-inspired neural architecture with unscented and extended Kalman filter optimization methods in order to model complex trajectory kinematics. Employing our previously developed robotic arm, we trained the device to track the trajectory with the perceptron model. The Kalman filter-trained perceptron model achieved prediction-correction process by adding weights to the corresponding synapses in the neurons attributing to error learning by induced plasticity as in neural microcircuits.