Design of a vision-based object tracker for the automated capture of objects by a manipulator

The trajectory of a tennis ball swinging on a pendulum can be estimated by means of object recognition by a stereo camera. Similarly, the motion of the ball can be modelled as a mathematical pendulum with damping. The extended Kalman filter provides the estimated state variables, the deflection angle and the angular velocity of the ball on the pendulum. The prediction of the trajectory of the ball then makes it possible to catch it autonomously at a suitable point by a manipulator. The trajectory for capturing the ball is implemented via a simple motion profile taking into account the temporal and spatial components. Experiments have shown that 85% of all attempts to catch the ball have been successful. Reasons for the unsuccessful attempts were sought and investigated in terms of both temporal and local displacements. Furthermore, improvements and extensions as well as the application of the implementation to other problems of the same methodological class are discussed. Copyright (c) 2022 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)