Large-Scale Multi-Robot Assembly Planning for Autonomous Manufacturing

TL;DR

This paper tackles scalable autonomous manufacturing with large robot fleets by introducing a full algorithmic stack that transforms CAD-like product specifications into distributed, collision-free assembly plans. It blends transport-unit configuration, radial staging layouts, a graph-repair MILP for task allocation, and a three-layer online collision-avoidance controller (Modified Tangent Bug, Prioritized Dispersion, and Generalized RVO), all validated with an open-source Julia-based simulator ConstructionBots.jl and demonstrations on LEGO-derived targets including a Saturn V with $1845$ parts, $306$ subassemblies, and $250$ robots in under three minutes on a standard laptop. The key contributions are (i) a scalable geometry-driven planning pipeline that decouples high-level planning from fine manipulation, (ii) a novel graph-repair MILP framework enabling collaborative transport task assignment, (iii) efficient radial staging and carrying-position heuristics, and (iv) a distributed execution strategy with robust deadlock-avoidance mechanisms. The approach enables rapid exploration of design trade-offs, supports customization and reconfiguration for Industry 4.0 settings, and provides a practical platform for researchers to study large-scale multi-robot assembly and coordination.

Abstract

Mobile autonomous robots have the potential to revolutionize manufacturing processes. However, employing large robot fleets in manufacturing requires addressing challenges including collision-free movement in a shared workspace, effective multi-robot collaboration to manipulate and transport large payloads, complex task allocation due to coupled manufacturing processes, and spatial planning for parallel assembly and transportation of nested subassemblies. We propose a full algorithmic stack for large-scale multi-robot assembly planning that addresses these challenges and can synthesize construction plans for complex assemblies with thousands of parts in a matter of minutes. Our approach takes in a CAD-like product specification and automatically plans a full-stack assembly procedure for a group of robots to manufacture the product. We propose an algorithmic stack that comprises: (i) an iterative radial layout optimization procedure to define a global staging layout for the manufacturing facility, (ii) a graph-repair mixed-integer program formulation and a modified greedy task allocation algorithm to optimally allocate robots and robot sub-teams to assembly and transport tasks, (iii) a geometric heuristic and a hill-climbing algorithm to plan collaborative carrying configurations of robot sub-teams, and (iv) a distributed control policy that enables robots to execute the assembly motion plan collision-free. We also present an open-source multi-robot manufacturing simulator implemented in Julia as a resource to the research community, to test our algorithms and to facilitate multi-robot manufacturing research more broadly. Our empirical results demonstrate the scalability and effectiveness of our approach by generating plans to manufacture a LEGO model of a Saturn V launch vehicle with 1845 parts, 306 subassemblies, and 250 robots in under three minutes on a standard laptop computer.

Figures (14)

• Figure 1: An overview of the proposed multi-robot assembly planning system. (a) Starting with a CAD-like project specification, the process evolves to determine (b) the configuration of transport teams, then calculates a (c) hierarchical geometric approximation of parts and transport units. (d) Based on the geometry, a spatial layout for the construction site is designed. (e) Task allocation and team formations are computed, following which (f) a decentralized strategy ensures collision-free execution. (g) The entire planned procedure is simulated, culminating in the (h) final assembled project. The arrows indicate the sequential flow of the planning process.
• Figure 2: A visualization of the project specification for our example tractor assembly. The final assembly is composed of three subassemblies and a few individual parts.
• Figure 3: A visualization of the transport unit configurations for several of the assemblies and objects associated with the tractor project. The payload geometry is shown in black and white. The convex hull is highlighted in green. Robots are shown as blue disks, with their carrying positions highlighted with smaller orange disks.
• Figure 4: Over-approximated geometries (sphere, vertical cylinder, octagonal prism) for the final tractor assembly (top row) and for a transport unit.
• Figure 5: An example solution to a radial layout optimization problem. Circles $1$, $2$ and $3$ are placed precisely at their respective desired orientations relative to the center circle. Circles $4$ and $5$, however, are forced to split the difference because they would overlap if placed at their desired orientations.