Highly adaptable optimal control of an autonomous insect walking robot in the real world

Abstract

It is important for robotics to realize an autonomous walking robot that can adapt itself to unpredictable changes of the environment in the real world. In order to control a robot to achieving a purpose, it is necessary to solve the inverse problem. Generally, the inverse problem becomes a kind of the ill-posed problem. Usually, this kind of problem has been solved by adding some constraints externally to make it a well-posed problem. In the real world, however, it is impossible to define such an objective function and to set constraints in advance, because the environment changes unpredictably. It means that the traditional control method based on causality can not apply to robot walking in the real world. On the contrary, animals seem to use available input for information retrieval by integrating different information. It is of great importance to clarify this mechanism to implement a novel information processing system, then it might provide a human user with collaborative activities with artificial systems. We propose a mechanism to solve the ill-posed problem and apply it to insect walking in an unpredictably changing environment. These mechanisms are quite new and applicable to other artificial control systems.