Reducing variation in agri-food production systems with an innovative control method
Robbert van der Kruk defended his PhD thesis at the Department of Mechanical Engineering on May 14th.
The food production industry accounts for more than 10% of the global manufacturing sector. This industry has had a low-tech reputation for a long time, but that鈥檚 changing rapidly. Robot technology has revolutionized the sector and even small companies have adopted new technologies. Though, variability in weight, size, color and ripeness of agri-food products poses challenges to these technologies. Robbert van der Kruk explores in his PhD research how control system technology can play a vital role in variation reduction in the agri-food production industry.
Food production systems are used to optimize deliveries while ensuring high quality and cleanliness. Variability in agri-food products significantly affect these systems and this makes it difficult to use robot technology. In food production lines, vacuum grippers are often used for robots to grasp food products. The challenge with these robotic pick-and-place tasks is that they suffer residual vibrations when suction cups are used and that limits productivity. Robbert van der Kruk introduces an innovative control method for suctioned products in robotic pick-and-place applications with his PhD research. The aim of this method is to reduce the peel-off force and overshooting.
Design approach
The input shaping control method that is this developed within this PhD research is based on a prefilter of the trajectory generator in the product coordinate system, the task space. Robbert van der Kruk outlined a design procedure based on a robot mechanism, gripper and product model containing the relationship between the relative overshoot and the vibration-to-acceleration time ratio, allowing for the highest acceleration value while maintaining accuracy. Next to that, this research describes that the trajectory prefilter can also significantly reduce residual vibrations in flexible joint or flexible link robotic systems caused by the reference trajectory. A comparison with high-order setpoint feedforward clarifies the differences. A design approach is presented to maximize acceleration while maintaining acceptable overshoot limits. Robustness against product variations such as mass and length for the input shapers is addressed in relation to overshoot reduction and overall motion time.
Seed production-inventory model
The insights from systematic literature reviews on production-inventory control systems provide a theoretical basis for the modelling and control of the inventory level on the variability of perennial crop seeds. The production planning and inventory control for perennial crop seeds is characterized by multi-year production and multiple growth cycles. A time-discrete model parameterized with historical data is developed within this PhD research and three control schemes are formulated: a classical feedback-feedforward PID controller, a feedback-feedforward PID controller with a Smith Predictor, and a Receding Horizon Optimal Control scheme (MPC). The goal was to present and validate a novel seed production-inventory model, representing multi-year seed production and extending the dead-time delay model without production level uncertainty. This method was implemented in a manual receding horizon optimization planning application module in an agricultural CRM system.
Practical implementation
The results from this research contribute to the practical implementation of vibration reduction and an increase in flexibility in robotic agri-food production systems and mechatronics systems in general with prefilter control methods. An example of the practical usage of receding-horizon optimal control in production-inventory systems for crop seeds is given to improve the delivery of crop seeds.
Title of PhD thesis: . Supervisors: Prof. Herman Bruyninckx and Prof. Ren茅 van de Molengraft.