Abstract
The maintenance of standing balance requires information on the environment if the environment contains uncertainty. From this point of view, we have proposed a learning scheme for locomotion pattern by use of the information on the ground reaction forces. The motion pattern is represented by the weighted sum of basis functions, where learning corresponds to the determination of their weights. However, there are no discussions on what kind of basis function is effective. In this paper, we select trigonometric function, polynomial function and radial basis function as families of the basis functions, and compared the learning result by computer simulations using each function family. As a result, the effectiveness of the trigonometric functions or radial basis functions is confirmed.
