Layered Cost-Map-Based Traffic Management for Multiple Automated Mobile Robots via a Data Distribution Service

Seungwoo Jeong, Taekwon Ga, Inhwan Jeong, Jongkyu Oh, Jongeun ChoiarXiv 2022 · DOI · OpenReview

TL;DR

This paper introduces a novel layered cost-map approach for managing multiple autonomous mobile robots (AMRs) in shared environments. By leveraging a Data Distribution Service (DDS) for communication and introducing specialized "filters" (prohibition, lane, fleet, and region), the system enables more robust and intuitive traffic control, preventing deadlocks and improving efficiency.

What Problem Does This Paper Address?

Managing multiple AMRs in dynamic environments, like factories or warehouses, presents significant challenges. Unlike traditional automated guided vehicles (AGVs) that follow fixed paths, AMRs are more flexible but lack inherent traffic management capabilities. This can lead to issues like:

  • Deadlocks: AMRs getting stuck when trying to navigate narrow passages simultaneously.
  • Reduced Efficiency: AMRs interfering with each other's tasks or operating sub-optimally due to unpredictable interactions.
  • Lack of Intuitive Control: Difficulty in setting up rules for robot behavior, such as restricted areas or one-way paths, without complex programming.

Existing methods often rely on complex path planning or separate network configurations, which can be cumbersome. This paper aims to provide a unified, flexible, and intuitive traffic management system for AMRs.

Methods

The core of the proposed solution is a layered cost map combined with a Data Distribution Service (DDS) for inter-robot communication.

  • Data Distribution Service (DDS): AMRs communicate by publishing and subscribing to "topics" within a shared DDS domain. This allows for efficient, data-centric sharing of information like robot positions, map data, and control commands. A central traffic management server coordinates these communications.
  • Layered Cost Map: Instead of a single cost map, the system uses multiple layers, each representing a specific traffic management aspect. These layers are combined using an OR operation to create a master cost map that guides AMR navigation. The key novel layers are:
  • Prohibition Filter: Allows users to define areas that AMRs must avoid, essentially creating "no-go" zones. This is implemented by assigning a high cost to these areas.
  • Lane Filter: Enforces directional traffic rules, such as one-way passages, by assigning costs based on the AMR's orientation relative to the desired lane.
  • Fleet Layer: Enables AMRs to share their positions with each other (via the traffic management server). This information is used to create a dynamic cost map around other AMRs, promoting collision avoidance.
  • Region Filter: Manages exclusive access to specific areas. An AMR can request a "ticket" to enter a designated region, preventing other AMRs from entering until the ticket is released. This is crucial for preventing deadlocks in narrow passages or for managing shared workspaces.

The system uses the Dynamic Window Approach (DWA) for both global and local planning, which is a computationally efficient algorithm well-suited for cost-map-based navigation.

Datasets

Not discussed in the paper. The experiments were conducted with real-world AMRs in a laboratory setting.

Key Insights and Ideas

  • Layered Cost Maps for Complex Scenarios: The idea of breaking down traffic management into distinct, combinable layers (prohibition, lane, fleet, region) is a powerful way to handle diverse and complex traffic rules in a unified framework.
  • DDS as a Communication Backbone: Using DDS simplifies inter-robot communication by providing a standardized, data-centric publish-subscribe mechanism, eliminating the need for complex custom network setups.
  • Intuitive User Control: The prohibition and region filters, in particular, offer a more intuitive way for users to define robot behavior and manage shared spaces compared to traditional programming methods.
  • Proactive Deadlock Prevention: The region filter is a key innovation for preventing deadlocks in narrow passages by ensuring only one AMR can pass at a time, managed through a ticket system.
  • Fleet Layer for Cooperative Awareness: Sharing AMR positions via the fleet layer allows for a form of "swarm intelligence" where robots are aware of each other's presence and can proactively avoid collisions.

Key Results

The paper demonstrates the effectiveness of each proposed filter through experimental validation with real-world Hyundai Robotics AMRs:

  • Prohibition Filter: Successfully prevented an AMR from entering a designated area (under a desk with legs), even when the lidar sensor only detected partial obstacles.
  • Lane Filter: Enabled AMRs to navigate around obstacles by enforcing a one-way direction, forcing them to take a specific route.
  • Fleet Layer: Demonstrated collision avoidance between two AMRs by dynamically adjusting cost maps based on shared positions.
  • Region Filter: Effectively managed narrow paths, allowing one AMR to pass with higher priority and then the other, and also managed exclusive working areas, ensuring only one AMR occupied a designated space at a time.

Qualitative results are presented through figures showing the AMRs' paths and the resulting cost maps under different filter configurations.

Limitations and Future Work

  • Scalability of the Central Server: While a single traffic management server is used, its scalability for a very large number of AMRs is not explicitly discussed.
  • Dynamic Environment Adaptation: The current system focuses on static or semi-static configurations of filters. Further work could explore more dynamic adaptation of these rules based on real-time environmental changes or task priorities.
  • Human-Robot Interaction: While the paper mentions human coexistence, the current filters are primarily for AMR-AMR interaction. Extending these to more sophisticated human-aware navigation could be a future direction.
  • Performance Metrics: While qualitative results are strong, detailed quantitative metrics on task completion time, collision rates, or efficiency gains under various scenarios could further strengthen the claims.

Summary updated

Abstract

This letter proposes traffic management for multiple automated mobile robots (AMRs) based on a layered cost map. Multiple AMRs communicate via a data distribution service (DDS), which is shared by topics in the same DDS domain. The cost of each layer is manipulated by topics. The traffic management server in the domain sends or receives topics to each of AMRs. Using the layered cost map, the new concept of prohibition filter, lane filter, fleet layer, and region filter are proposed and implemented. The prohibition filter can help a user set an area that would prohibit an AMR from trespassing. The lane filter can help set one-way directions based on an angle image. The fleet layer can help AMRs share their locations via the traffic management server. The region filter requests for or receives an exclusive area, which can be occupied by only one AMR, from the traffic management server. All the layers are experimentally validated with real-world AMRs. Each area can be configured with user-defined images or text-based parameter files.
BibTeX
@misc{jeong2022layered,
  author = {Seungwoo Jeong and Taekwon Ga and Inhwan Jeong and Jongkyu Oh and Jongeun Choi},
  title = {{Layered Cost-Map-Based Traffic Management for Multiple Automated Mobile Robots via a Data Distribution Service}},
  year = {2022},
  eprint = {2207.08902},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2207.08902},
}
Figure