Automatic control for swing-free control of suctioned products in robotic pick-and-place operations

摘要

This paper introduces a new control method for suctioned products in robotic pick-and-place applications with the aim of significantly reducing the peel-off force and overshooting. A controlled cart-pendulum-mass model approximates the complex dynamics of the robot with the attached load. A design procedure is outlined based on the relationship between relative dynamic placement inaccuracy and the ratio of vibration time to move time. An algorithm is outlined for the implementation of the design method. This procedure enables the determination of the highest acceleration value that maintains accuracy within the desired range. This paper builds on previously published conference work (van der Kruk et al., 2023), incorporating a self-adjusting method, a specialised gripper, validation for significant mass variations, a frequency domain analysis and experimental results using chicken fillets handling. A passive rotational single degree of freedom is incorporated into a standard bellow vacuum gripper to minimise the torque applied to the end effector alongside input shaping. The accelerations allowed in the end effector are increased. Robustness against peel-off at higher accelerations is improved while leveraging the advantages of swing-free controls. Automatic tuning of the control parameters is achieved through a background subtraction method that measures the tracking errors in the joint motion controllers. The effectiveness of the proposed method is validated through a practical use case involving chicken fillet packaging with a fast industrial delta robot. This is a weight and pick-up spread case test. All fillets in the industrial use-case weight distribution range can be consistently picked and handled using the same constant parameters without experiencing peel-off. The Zero Vibration input shaper demonstrates the best overall performance for realistic variations (such as weight and pick-up pose), well within 10% of the variations of the natural frequency. • Insights into the vibration and peel-off behaviour of products during pick-and-place movement are derived. • Solutions are explored to improve the input-shaping method for unknown products by robustness against weight variations, using a passive additional one degree of freedom to the end-effector. • When products cannot be processed with constant input shape parameters, in situ measurements could be performed during the pick-and-place movement to find the natural frequency and damping ratio. • The most promising method for product identification is to derive the natural frequency during transfer movement. This can be done by analysing the error signal in the motor drives or incorporating a sensor into the hinge vacuum gripper. • For both methods, the natural frequency may be determined during the first half of the horizontal movement and input shaping is applied in the second half of the movement. • The majority of relevant research of swinging objects focusses on cranes rather than industrial robots. • Instead of using more expensive hardware or imitation learning methods, these robots can now reduce vibrations using a straightforward software technique.