Model-based control systems are among most efficient and reliable robotic control systems. However, extracting manipulator’s dynamic modelling is challenging. In this course project, a one-layer MLP neural network is used to model the lumped function H, which allows for estimation of joint torques based on current joint states. The model is investigated using RMSE and NRMSE metrices. It, also, is further investigated by comparing the performance with MATLAB MLP and RBF modellings that we previously discussed in the course. Compare to them, the model shows good results. However, generally speaking the model performance does not seam trustable for control implementation, and there is a need for upgrading and improving model performance before implementation.
If you have any questions or need further assistance, feel free to contact the project maintainers:
Faezeh Haghverd (haghverd@ualberta.ca)