Task prioritization for automated robotic item picking

This paper describes a task scheduling method designed for a robotic picking system. The main goal of the system is to fulfill an order, ie, pick all the items in an order list and place them into the order bin, as fast as possible with the least number of unfulfilled items or errors. Picking an item on the list is considered as one task. The system will prioritize and schedule first the tasks with higher chances of being executed successfully. The probability for successful pick is computed from the vision data and grasping information. The items with higher chances of failure to be picked are pushed to the end of the queue. Forty experiments were conducted using randomly generated arrangement of items in a shelf and a randomly generated task order list. The test result shows that the strategy delivers an average fulfillment rate of 90% and an average fulfillment efficiency rate of 71%. The strategy described in this paper could be used to determine items that could be picked with human assistance instead of by robot alone.