From Simulation to Real-World Robotic Mobile Fulfillment Systems

Lin Xie, Nils Thieme, Hanyi Li

TL;DR

This paper bridges the gap between simulating Robotic Mobile Fulfillment Systems (RMFS) and deploying them in the real world. They introduce a four-step process, starting with adapting a simulator into core software, and culminating in testing algorithms on actual robots via a novel "XOR-bench" platform. This work is crucial for validating RMFS research and enabling practical deployment of these automated warehouse systems.

What Problem Does This Paper Address?

Robotic Mobile Fulfillment Systems (RMFS) are an emerging class of automated warehouses where robots transport shelves (pods) to human operators for picking and replenishment. While simulation frameworks like RAWSim-O exist to study the complex, interdependent decision problems within RMFS (e.g., robot task allocation, pod selection), a significant challenge remains: how to ensure that algorithms developed and tested in simulation are truly applicable and effective in real-world RMFS. The paper addresses this "simulation-to-reality gap" by proposing and demonstrating a structured methodology for transitioning from simulation to a functional, real-world RMFS.

Methods

The authors propose a four-step process to bridge the simulation-to-reality gap:

  1. Simulator to Core Software: The RAWSim-O simulator is adapted into a core RMFS software. This involves implementing an Application Programming Interface (API) for integration with external systems like ERPs and an agent interface for communication with robots and station applications.
  2. Connecting with ERP and Station Apps: The RMFS core software is connected to an open-source Enterprise Resource Planning (ERP) system, Odoo, and custom-developed station applications. This allows for real-time inventory management and task assignment based on customer orders.
  3. Integration with Mini-Robots (XOR-bench): A novel "XOR-bench" platform is designed. This teleoperated environment allows for the integration of the RMFS core software with mini-robots. The XOR-bench provides functionalities like live video streaming, program download to robots, and computer vision-based robot positioning.
  4. Testing and Validation on Mobile Industrial Robots: The XOR-bench is further utilized to test and validate algorithms on actual mobile industrial robots. This step leverages the platform's internet connectivity for remote experimentation.

Datasets

Not discussed in the paper. The paper focuses on the methodology and system integration rather than specific datasets for training or evaluation.

Key Insights and Ideas

  • Bridging the Simulation-to-Reality Gap is Crucial: The paper highlights that without a clear path from simulation to real-world deployment, research in RMFS risks remaining theoretical and not translating into practical solutions.
  • The XOR-bench as a Key Enabler: The custom-designed XOR-bench is a significant contribution, providing a flexible and accessible platform for integrating RMFS software with both mini-robots and industrial robots, facilitating research and education.
  • Modular Integration is Key: The four-step process emphasizes a modular approach, allowing for incremental development and testing, from adapting simulation to connecting with enterprise systems and finally to real robot deployment.
  • ERP Integration is Essential for Real-World Functionality: Connecting the RMFS core software with an ERP system like Odoo is vital for managing inventory, orders, and transfers, making the system operational in a warehouse context.
  • Remote Experimentation is Feasible and Beneficial: The paper demonstrates the effectiveness of remote testing of RMFS algorithms on industrial robots, reducing the cost and complexity associated with physical experimentation.

Key Results

The paper successfully demonstrates the feasibility of its four-step process. They show:

  • The adaptation of the RAWSim-O simulator into a functional RMFS core software.
  • Successful integration of the RMFS core software with the Odoo ERP system and custom station applications.
  • The XOR-bench platform effectively facilitates the integration and testing of RMFS software with mini-robots (LEGO Mindstorms EV3).
  • The system architecture allows for remote testing and validation of algorithms on mobile industrial robots, showcasing the potential for widespread accessibility and collaboration.

Specific quantitative results on algorithm performance (e.g., path planning efficiency) are not the primary focus, but the successful integration and demonstration of the system's capabilities are the main outcomes.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the process of integrating RMFS core software with mini-robots and industrial robots is complex and requires interdisciplinary work.
  • While the XOR-bench is designed for RMFS, its current implementation is focused on specific experimental setups.
  • The paper mentions that for the remote experiment with industrial robots, they did not need the video-based positioning system of the XOR-bench due to the robot's built-in capabilities. This implies that the XOR-bench's vision system might be more suited for the mini-robots.
  • Future Work:
    • The authors suggest that the XOR-bench can be used in the future to test other types of automated warehousing systems, such as those where humans and robots work collaboratively in shared spaces.
  • The developed process flow and XOR-bench platform are beneficial for testing other types of automated warehousing systems.
  • The paper implies that further development and refinement of the XOR-bench and the integration process will continue to enhance RMFS research and deployment.

Summary updated

Abstract

In a new type of automated parts-to-picker warehouse system – a Robotic Mobile Fulfillment System (RMFS) – robots are sent to transport pods (movable shelves) to human operators at stations to pick/put items from/to pods. There are many operational decision problems in such a system, and some of them are interdependent and influence each other. In order to analyze the decision problems and the relationships between them, there are two open-source simulation frameworks in the literature, Alphabet Soup and RAWSim-O. However, the steps between simulation and real-world RMFS are not clear in the literature. Therefore, this paper aims to bridge this gap. The simulator is firstly transferred as core software. The core software is connected with an open-source ERP system, called Odoo, while it is also connected with real robots and stations through an XOR-bench. The XOR-bench enables the RMFS to be integrated with several mini-robots and mobile industrial robots in (removed) experiments for the purpose of research and education.
BibTeX
@inproceedings{xie2018from,
  author = {Lin Xie and Nils Thieme and Hanyi Li},
  title = {{From Simulation to Real-World Robotic Mobile Fulfillment Systems}},
  booktitle = {Logist. Res},
  year = {2018},
  doi = {10.23773/2019\_9},
}
Figure