[Paper Review] Toward Fast and Optimal Robotic Pick-and-Place on a Moving Conveyor

Shuai D. Han, Si Wei Feng+1 more|arXiv (Cornell University)|Dec 17, 2019

Robotic Path Planning Algorithms28 references4 citations

TL;DR

This paper proposes dynamic programming-based algorithms—OptSeqDP and SubOptDP—that compute time-optimal picking sequences for robotic pick-and-place on moving conveyors, significantly outperforming greedy heuristics. The methods achieve 10–40% higher throughput by optimizing object order under geometric and dynamic constraints, with sub-second runtime for hundreds of objects.

ABSTRACT

Robotic pick-and-place (PnP) operations on moving conveyors find a wide range of industrial applications. In practice, simple greedy heuristics (e.g., prioritization based on the time to process a single object) are applied that achieve reasonable efficiency. We show analytically that, under a simplified telescoping robot model, these greedy approaches do not ensure time optimality of PnP operations. To address the shortcomings of classical solutions, we develop algorithms that compute optimal object picking sequences for a predetermined finite horizon. Employing dynamic programming techniques and additional heuristics, our methods scale to up to tens to hundreds of objects. In particular, the fast algorithms we develop come with running time guarantees, making them suitable for real-time PnP applications demanding high throughput. Extensive evaluation of our algorithmic solution over dominant industrial PnP robots used in real-world applications, i.e., Delta robots and Selective Compliance Assembly Robot Arm (SCARA) robots, shows that a typical efficiency gain of around 10-40% over greedy approaches can be realized.

Motivation & Objective

• To address the sub-optimality of widely used greedy heuristics (e.g., FIFO, SPT) in robotic pick-and-place on moving conveyors.
• To develop a principled, computationally efficient algorithmic framework that guarantees time-optimal or near-optimal picking sequences for finite horizons.
• To evaluate the performance of the proposed algorithms on realistic industrial robot models (Delta and SCARA) under varying object densities and conveyor speeds.
• To demonstrate that optimal sequencing can yield substantial real-world efficiency gains in high-throughput industrial automation settings.

Proposed method

• Formalize the robotic PnP problem on a moving conveyor as a finite-horizon optimization task with geometric and dynamic constraints.
• Develop the OptSeq algorithm as an exhaustive dynamic programming solution that computes exact optimal sequences for up to 20 objects.
• Introduce OptSeqDP and SubOptDP by applying dynamic programming with pruning and locality-based heuristics to scale to 100+ objects.
• Integrate robot-specific kinematics and dynamics into the cost function to model realistic motion time profiles.
• Use simulation environments with telescoping, Delta, and SCARA robot models to evaluate algorithmic performance.
• Apply performance metrics such as total number of successfully picked objects and execution time efficiency across varying object densities.

Experimental results

Research questions

• RQ1 How sub-optimal are standard greedy heuristics (e.g., FIFO, SPT) in terms of time efficiency for robotic pick-and-place on moving conveyors?
• RQ2 Can dynamic programming be effectively applied to compute optimal or near-optimal picking sequences under realistic robot dynamics and geometric constraints?
• RQ3 To what extent can the proposed algorithms scale to hundreds of objects while maintaining real-time feasibility?
• RQ4 What is the performance gain of the proposed algorithms over existing heuristics in continuous, high-throughput PnP operations?
• RQ5 How does the performance of the proposed algorithms vary across different robot types (e.g., Delta, SCARA) and object distributions?

Key findings

• Greedy heuristics like FIFO and SPT can lead to up to 40% loss in time efficiency compared to optimal solutions, as analytically demonstrated on a simplified telescoping robot model.
• The OptSeqDP and SubOptDP algorithms compute optimal or near-optimal picking sequences in under one second for up to 100 objects, enabling real-time deployment.
• In one-shot scenarios, the proposed algorithms achieve up to 20% reduction in execution time compared to FIFO, Euclidean, and SPT heuristics.
• In continuous conveyor operations, the proposed methods yield a consistent 10–40% improvement in the number of successfully picked objects over greedy baselines.
• The performance gain is robust across different robot models (telescoping, Delta, SCARA) and object distributions (Poisson and uniform).
• The algorithms show significant advantage in dense object scenarios where greedy methods suffer from poor object selection due to long travel times.