The packaging of pre-packaged meal salads available in supermarkets is currently only partially automated. Apart from the main source of carbohydrates (often some form of pasta or potatoes) and/or lettuce, the ingredients are added by hand. This is done by employees positioned at an assembly line, who take the ingredients from a crate by hand and drop them into the meal. The precision of this method is limited, given the inability of humans to accurately estimate weight. Furthermore, this task is very repetitive and menial, making it difficult for employees to stay focused and deliver consistent portions of ingredients during the full duration of their shift.
The aim of this project is to determine whether a 6DOF robotic arm with a scoop as an end-effector can form a suitable replacement for human workers in ready-made meal factories. To this end, a parametrised scooping motion was devised. Depth and distance parameters were varied, gathering approximately 1500 datapoints on the relation between parameters and gathered mass for three ingredients (bellpepper cubes, dried macaroni & canned sweetcorn). Furthermore, incremental dispensing of the ingredient is investigated to see if this leads to an increase in accuracy and consistency in mass.
Results show that the performance of the robotic scooping motion is heavily dependent on ingredient characteristics such as uniformity in size and shape, compressibility and surface friction. Incremental dispensing poses the highest potential for robotic scooping for bell pepper cubes and canned sweetcorn, while varying the scoop parameters yields the best results for dried macaroni. However, manual scooping remains more consistent than robotic scooping, especially for a fixed target mass.