Automated Generation of Continuous-Space Roadmaps for Routing Mobile Robot Fleets

TL;DR

This work addresses the challenge of routing large fleets of mobile robots in intralogistics by generating continuous-space roadmaps that balance redundancy and query efficiency. It introduces a four-step pipeline that discretizes free space around stations and environmental features, constructs bidirectional edges under minimum distance constraints, and optimizes the resulting graph using Yen’s $K$-shortest paths with dynamic edge costs before smoothing trajectories with cubic splines. Key contributions include explicit minimum inter-node and node-edge distance constraints, transport-demand integration via a $T$ matrix, planarity-aware pruning, and a smoothing stage that yields feasible, efficient routes. Across three intralogistics environments, the method yields roadmaps with lower structural complexity, higher redundancy, and near-optimal path lengths (normalized mean shortest path around $1.03$–$1.046$), outperforming 4-connected, 8-connected, and random baselines and enabling scalable, robust fleet routing.

Abstract

Efficient routing of mobile robot fleets is crucial in intralogistics, where delays and deadlocks can substantially reduce system throughput. Roadmap design, specifying feasible transport routes, directly affects fleet coordination and computational performance. Existing approaches are either grid-based, compromising geometric precision, or continuous-space approaches that disregard practical constraints. This paper presents an automated roadmap generation approach that bridges this gap by operating in continuous-space, integrating station-to-station transport demand and enforcing minimum distance constraints for nodes and edges. By combining free space discretization, transport demand-driven $K$-shortest-path optimization, and path smoothing, the approach produces roadmaps tailored to intralogistics applications. Evaluation across multiple intralogistics use cases demonstrates that the proposed approach consistently outperforms established baselines (4-connected grid, 8-connected grid, and random sampling), achieving lower structural complexity, higher redundancy, and near-optimal path lengths, enabling efficient and robust routing of mobile robot fleets.

Figures (6)

• Figure 1: Roadmap of an abstract environment. Station interaction points are connected by a roadmap consisting of nodes and edges. The roadmap adapts to environmental boundaries around obstacles (gray areas), which are geometrically expanded (light gray areas), while providing redundant, continuous-space paths within free space (white areas).
• Figure 2: Minimum distance constraints for roadmap nodes and edges based on mobile robot dimensions. $d_{V\text{min}}$ denotes the minimum inter-node distance, and $d_{VE\text{min}}$ is the minimum node-edge distance.
• Figure 3: Automated roadmap generation process shown for an abstract environment. (a) Visibility graph connecting the interaction points of the environment for a uniform transport demand, with added (blue crosse) and not added nodes (green squares). (b) Full roadmap after defining local grids and edge connections. (c) Reduced roadmap after pruning nodes and edges to adjust the redundancy. (d) Optimized roadmap after removing crossing edges and straightening paths.
• Figure 4: Optimized roadmap with smooth trajectories. Blue lines show the smoothed roadmap generated using cubic splines, respecting the allowed deviation margin $d_{ad}$ from nodes, represented by the green dotted circle.
• Figure 5: Intralogistics environments used for evaluation. Red areas are stations with their interaction points marked by red rhombuses. The optimized roadmaps using the proposed approach are visualized by the blue crosses and blue dashed lines.