Robotic box pushing under indeterminate anisotropic friction properties

Manipulation and secure transportation of loads by robotic manipulators along predetermined paths is still an open challenge in the robotics field. In this article, the problem of pushing and driving an object on a surface employing a robot to reach a specific destination is considered under a-priori unknown friction conditions. For this purpose, the equations of motion are derived according to the limit surface conceptualization of friction force. Although friction parameters may be considered locally constant, nonetheless they gradually vary over large surfaces. The friction term is thus faced with a slowly time-varying indeterminacy and needs an online estimation scheme. By defining a mission consisting of pushing an object via a robot to find a path among obstacles, an optimum path is firstly planned then a non-linear model predictive controller is used to track the desired path. In order to alleviate the indeterminacy due to friction, the corresponding terms in the equations of motion are set as a disturbance which is estimated by a purposely designed observer and then eliminated in a feedforward manner. There is also no need to perform preliminary offline tests.