Graph-Based Adaptive Planning for Coordinated Dual-Arm Robotic Disassembly of Electronic Devices (eGRAP)

TL;DR

The paper addresses the challenges of automated e-waste disassembly by introducing eGRAP, a perception-driven, graph-based planning framework that coordinates two robot arms. It integrates live vision outputs into a directed Part Graph with precedence and access constraints, uses topological sequencing to generate feasible action sets, and assigns actions to two arms under collision and workspace constraints with online plan updates. Key contributions include online graph maintenance from live detections, a dual-arm scheduling scheme that enables both hold–operate and independent tasks, and a device-agnostic loop that generalizes across products by updating detectors, rules, and action templates. Experiments on 3.5-inch HDDs from three manufacturers show full teardowns completed within about 22 minutes on average, with high perception accuracy and robust, adaptive coordination. This work advances scalable, autonomous disassembly for sustainable manufacturing and recycling pipelines.

Abstract

E-waste is growing rapidly while recycling rates remain low. We propose an electronic-device Graph-based Adaptive Planning (eGRAP) that integrates vision, dynamic planning, and dual-arm execution for autonomous disassembly. A camera-equipped arm identifies parts and estimates their poses, and a directed graph encodes which parts must be removed first. A scheduler uses topological ordering of this graph to select valid next steps and assign them to two robot arms, allowing independent tasks to run in parallel. One arm carries a screwdriver (with an eye-in-hand depth camera) and the other holds or handles components. We demonstrate eGRAP on 3.5in hard drives: as parts are unscrewed and removed, the system updates its graph and plan online. Experiments show consistent full disassembly of each HDD, with high success rates and efficient cycle times, illustrating the method's ability to adaptively coordinate dual-arm tasks in real time.

Figures (7)

• Figure 1: Dual-Arm testbed with a Manipulation arm (vacuum gripper) and a Tooling arm (screw-driving tool, RGB–D camera, and micro-camera). A 3.5in HDD is fixed on a passive holder at the workspace centre.
• Figure 2: Block diagram of the eGRAP framework. Perception outputs labeled parts and 3D poses. The Graph block encodes precedence and access rules. The Sequence Generator topologically orders ready parts and instantiates primitive actions. The Scheduler assigns actions to two arms and updates the plan online from new observations. In the timeline, blue outlines denote the tooling (screwdriver) arm and green outlines denote the manipulation (gripper) arm; fill colors denote action types: yellow = unscrew, purple = remove, orange = drop.
• Figure 3: Custom end-effector with a Soleilwear precision electric screwdriver, a 2D RGB micro-camera, two micro-servos for forward/reverse, and a controller. A Torx T8 bit is used.
• Figure 4: HDD brands and fasteners.
• Figure 5: HDD sequence generation with eGRAP (L1 example). Top: the detector finds screw and lid. Middle: detections become nodes; class rules add edges (e.g., screw$\rightarrow$lid). Bottom: a topological sort yields a valid order for lid removal. L2 and L3 follow the same pattern with different labels and rules.