The development of an efficient scheduling heuristic for multi-robot pick-and-place operations

As awareness of the ecological impact of material production and the demand for resources increases, recycling and reusing materials has become increasingly critical. The development of advanced sorting systems is essential to reduce the ecological footprint of producing new materials. This study expands the focus from hardware alone to include real-time intelligent scheduling algorithms, with the aim of improving the efficiency of material sorting systems. It explores optimisation algorithms tailored for the pick-and-place (PnP) sorting of recyclables, specifically addressing the complex time-dependent pick-and-place scheduling problem with time windows (TD-PnP-TW) on a moving conveyor belt. This problem involves sequencing object picking and assigning tasks to multiple robots to optimise system performance. The architecture of the system features stationary robots alongside a conveyor belt, each with a fixed workspace. An exact model is introduced, and in light of the problem's complexity and the limited time between computer vision-based object detection and robotic handling, a collaborative multi-robot scheduling heuristic is proposed. This heuristic dynamically considers the state of available robots while planning picking sequences. Experimental validation demonstrates that this algorithm improves system performance by over 4% compared to the best traditional decision rules, highlighting its effectiveness in optimising multi-robot picking systems.