Non-contact manipulation of a single solid object on a plane surface using multiple air jets (Experiments with fuzzy control)

We have been studying an air-jet manipulation technique for controlling the position and posture of a single solid object on a plane surface by manipulating the air flow and angles of multiple air jets. This technique has a highly nonlinear nature caused by the diffusivity of air jets, so that accurate plant modeling is very difficult. Therefore, when the model greatly differs from reality, a problem occurs where its control performance greatly decreases. In this research, the above problem was resolved by rule-based approach, not by model-based one. More specifically, position error, velocity and distance between the nozzle and object were defined as input variables. Individual nozzle air flow was defined as the output variable, and each fuzzy membership function was determined based upon experiment results found by the conventional method. With attention being focused on the specific conditions of “start of moving object,” “approaching the goal,” “moving within the goal area,” “separation from the goal,” “stopping immediately before the goal area” and “stopping within the goal” as the relative positional relationship between the goal and the object, and a fuzzy rule was designed for each condition. An experiment to manipulate a spherical object with four air jets was conducted and the effectiveness of the proposed method was proved through comparison with the conventional model-based method.