Robotisch manipuleren en sorteren van afvalstromen

Technological, social and economic advancements have led to new mass production and production technologies which resulted in products and packaging that are often harder to repair or recycle. To prevent depletion and material quality degradation, sorting and recycling are required. In an effort towards improving automated sorting processes, this work focuses on two aspects of robotic waste sorting: first, the selection and development of robotic gripper technologies for the manipulation of waste materials are investigated, and second, the scheduling of the robot sorting process. These aspects are implemented and evaluated within two industrial use cases. One, focussing on the sorting of metal scrap originating from the automotive and construction sector. The other, focussing on the manipulation and sorting of plastics and packaging waste. The first aspect of this work started with gripper selection. An elaborate literature review was performed to evaluate the state of the art in gripper technologies. This review is performed specifically for the manipulation of waste and the identified technologies are evaluated based on their technological characteristics, technological readiness, safety and material compatibility. Additionally, methods to support the actual gripper selection and design are proposed and demonstrated. After identifying potential technology gaps and interesting gripper technologies, three gripper technologies are further developed. The first gripper is a two fingered impact gripper. Within the development, different aspects of this gripper type are investigated, tested and validated. For example, different designs with different finger types, bar mechanism, pneumatic actuators and an electric actuator with the use of tendons are made. The High Airflow Vertical COnveying (HAVCO) gripper is the second gripper developed within this work. This gripper went through many iterations in the application of both use cases and had demonstrated high potential, specifically for the manipulation of small and lightweight objects. Due to the design of this novel gripper principle, the gripper eliminates part of the pick and place motion which is required when objects are removed with more traditional gripper technologies. This results in a pick-to-pick motion instead of a pick and place motion, which is more time efficient and enables the system equipped with the HAVCO gripper to have a higher pick rate. Additionally, the gripper doesn't require a precise grasping pose, simplifying integration of the gripper. Within the metal sorting use case, an early prototype of the HAVCO gripper reached up to 92,4% grasp success ratio with an average picking time of 0.32 seconds per pick. Within the plastic and packaging waste (PPW) sorting use case, the final prototype reached a success ratio between 60,3 and 100% with an average success ration over all the different type of PPW objects of 79,75%. The last developed gripper within this work is a novel gripper for the manipulation of metal objects. The gripper is specially developed for the metal scrap sorting use case. Even though the gripper principle showed potential, further testing of this gripper principle is postponed since a testing installation with a higher load capacity is required to further validate the influence of the process parameters. The second aspect of this work focuses on the optimization of multi-robot multi-gripper sorting systems by the integration of a novel scheduling heuristic. Two applications of the scheduling heuristic are evaluated. First, the integration of the scheduling heuristic within the sorting system to schedule the grasping of objects in the operating system is investigated. The heuristic attempts to improve the yield of the sorting system by grasping the objects with the specific robot-gripper combination with the highest grasping success rate and the lowest picking time. To evaluate the scheduling heuristics, a simulation environment is used. The impact of the robot type, conveyor belt velocity, waste stream composition, specific gripper set and waste-type-to-bin allocation has been evaluated within the simulation environment, indicating the impact of the proposed design decision. Within the first application, a novel heuristic is proposed which improved the overall system throughput by 4% over the best performing decision rule by scheduling the picks within the sorting system. Subsequently, this novel heuristic is adapted with the specific focus on the waste sorting industry. The modifications enable the heuristic to consider the varying efficiencies of different grippers. Due to the adaptation, the grasp success could be improved with up to 3,6% and the pick rate could be increased with up to 25,9% depending on the system layout. Additionally, the impact of the conveyor velocity in combination with the proposed heuristic is assessed. Increasing the conveyor velocity provides more material to the sorting system which is resulting in more opportunities for optimization. Increasing the conveyor belt velocity increases the grasp success rate with 2,9%, while the pick rate increased up to 37,5% with the higher conveyor belt velocity. To finalize the work, the three aspects are further discussed within the valorization chapter with the specific focus on the exploitation of the methods and technologies developed within this work.