Scheduling of Pick-and-Place Tasks in Shared Workspaces Using a Trajectory-Free Approximation Method

The scheduling of pick-and-place tasks in systems with multiple robots in shared workspaces is a well-studied topic with significant industrial relevance. We address a complex scenario involving four heterogeneous robot arms operating within a limited shared workspace, which was introduced by El Ghazi et al. (2024). Their approach leverages mixed-integer linear programming (MILP) modeling using trajectory simulations, achieving low makespans but at the cost of high computational time and model complexity. To improve efficiency, we propose an alternative approximation method based only on the durations of the trajectories and the relative positions of robots and objects, without prior knowledge of the space taken by the robots in their trajectory, therefore bypassing the need for spatial trajectory simulation. Our model incorporates hitboxes to approximate the space occupied by robots during tasks, therefore predicting collisions between robots, while reducing computational load. Experimental comparisons indicate that our method is significantly faster than the other approach. For larger instances (ten objects to pick up), our model often outperforms El Ghazi et al.’s model in terms of makespan.