The State of Robot Fleet Management Systems from an intralogistics perspective. Part 1: Introduction and Literature overview

This is the first part of a publication “The State of Robot Fleet Management Systems from an intralogistics perspective: Technologies, Challenges, Trends”.
In case you found the material interesting and you want to read the rest (with different readiness levels I have draft sections for drivers&enablers, emerging technologies, barriers&challenges, trends) — reach me via LinkedIn direct messages. Or if you want to become a co-author and make together a journal/conference paper based on that, send a direct message too.

Part 2: Drivers&Enablers
Part 3: Emerging Technologies

Introduction

Robot fleet management system (RFMS) primarily concerns itself with managing a group of vehicles to meet the goals and objectives obtained from an enterprise computer system [1]. As a means of vehicles in this publication, we consider Autonomous Mobile Robots (AMR) which in general are an advanced form of Automated Guided Vehicles (AGV). The difference between them is the capability of AMR to navigate autonomously using flexible paths without external physical elements such as painted or magnetic lines needed for AGV and avoid obstacles in a dynamic environment [2]. We examine RMFS in the context of an intralogistics environment, which is organization, realization, and optimization of internal material flow and logistic technologies as well as the goods transshipment in industry, trade, and in public institutions by means of technical components, partial and full systems and services [3].

Automation is proliferating in E-commerce and Industry 4.0, also increasing AMR adoption and application [4][5][6]. Logistics in these environments though is well regulated by rules, yet it leads to the complex behavior of the entire system [7]. To make robots operation efficient a RFMS is needed to tackle tasks such as job allocations, traffic management, integration with an operating environment, communication, interaction with a user. Though, still, there is no consensus or common awareness of the problem as well as no standards [8][9]. To address this problem, in this paper we define the current state of RFMS as a research topic including drivers, enablers, emerging technologies, barriers, challenges, and key research directions. Our findings will help researchers to make decisions on research topics and developers to define requirements and make the development process more straightforward.

Literature overview

Since the pivotal publications in the AGV fleet management field by Wurman et al [10] and Guizzo [11] on the Kiva system, many efforts had been spent to expand the knowledge that can be found in the comprehensive review by Oyekanlu et al [12]. Though there are not many studies that cover AMR-specific fleet management.

Andreasson et al [13] addressed the problem of realizing a complete efficient system for automated management of fleets of autonomous ground vehicles in industrial sites. A set of core requirements for systems comprising a fleet of AGVs is described. Within the SAUNA project, these requirements have been addressed within an integrated approach. A modular approach is proposed based on least commitment, which integrates perception, task allocation, motion planning, coordination, collision prediction, and control through a uniform constraint-based paradigm.

Mansouri et al [14] proposed to cast the problem of autonomous fleet management to a meta-CSP that integrates task allocation, coordination, and motion planning. A solution to the management of fleets of autonomous vehicles is proposed that (1) incorporates task allocation (2) accounts for individual feasible motions of each vehicle with respect to the vehicle’s kinematic model; (3) avoids deadlocks and collisions; (4) is able to accommodate externally imposed temporal and spatial constraints that are either metric or qualitative. The spatial and temporal dimensions of a task are represented as trajectory envelopes. These are collections of spatial and temporal constraints on vehicle trajectories and temporal constraints on when the task is to be achieved. The approach is grounded on a common constraint-based representation which is shared among various reasoners. This common representation includes both metric and qualitative knowledge enabling reasoning at both levels. The constraints capture heterogeneous requirements of a wide category of industrial domains. The core algorithm is a meta-CSP search that interleaves reasoning to achieve a feasible solution with respect to all requirements.

Yao et al [15] presented an approach for a Smart AGV Management System (SAMS), which combines real-time data analysis and digital twin models that can be deployed within complex manufacturing environments for optimized scheduling. SAMS collect real-time manufacturing operations data such as throughput rate, availability, and utilization of both production stations and AGVs to predict and optimize a schedule for material delivery. The key novelty proposed in this paper is the data analysis and connectivity with digital twins. SAMS consists of six main modules: communication module, data storage module, data processing module, Digital Twin module, decision-making module, and human-machine interface module. The use case shows the integration of the Smart AGV Manager based control system integrating with IoT devices, AGV, and manual assembly process.

Lutz et al [16] proposed a distributed approach for autonomous robots forming a fleet capable of performing transportation tasks within an industrial production system. The system is able to adapt to changes in the production flow by changing the software configuration only. The components on each robot are orchestrated by a Sequencer component coordinating task execution as modeled with the Domain Specific Language (DSL) SmartTCL. The coordination of the robots within the fleet can be separated into two parts, the coordination of the navigation and the coordination of the jobs and robots themselves. The components required for coordination could either run on a fixed infrastructure computer or on one of the robots in the fleet. By receiving jobs and sending back answers, the Sequencer realizes, on a symbolic level, the sparse interface between the fleet and individual robots. Keeping the robots autonomous and the interface sparse following the subsidiarity principle (resolving issues locally), increases the robustness and helps to master the overall complexity, not exposing unnecessary low-level details. The system can be adapted to changes in the production flow quickly by changing the software configuration only.

Čech et al [17] created a knowledge base for the design, selection, and implementation of AMR technology in the form of unit load carriers for supplying assembly lines in the automotive industry. The article presents a general procedure for a feasibility study of the project of AMR assembly line supply and a set of recommendations divided into five key categories: Technology, Management, Economics, Capacity, and Vendors. The basic research tool was a case study aimed at assessing the feasibility of introducing AMR technology for supplying a car assembly line with parts with irregular consumption in a leading passenger car manufacturer. Based on the performed research and case study, the general procedure for a feasibility study of the project of AMR assembly line supply has been proposed. The procedure includes eight steps in the recommended sequence, while some of the steps can be carried out in parallel.

Chitic et al [18] discussed the need of having a family of middlewares in large-scale multi-robot systems and how it facilitates software development. A comparison of the seven robotic frameworks presented above from a software engineering point of view because the middleware concept first emerged from this area and there is a lot of knowledge that can be transferred into robotic applications. We have grouped the comparative criteria into two major groups: Architecture and Infrastructure. Each major group is composed of different criteria relevant to the group. The Architecture evaluates the impact that the framework has over the host operating system and is composed of: Vendor Locking, Durable Data, Storage Services, Robustness to Failures. The Infrastructure evaluates tools and APIs provided by the middleware and is composed of Management and Monitoring, Multi-Robot Coordination Services, Communication. Concluded that a fleet of robots can be organized as a cloud of robots and not a cloud for robots. There is a lot of work in multi-robot systems to automatically distribute tasks of a greater problem. Our vision is to bring all the benefits of a cloud environment into the robotic fleet by allowing robots to share information, resources, computation power across heterogeneous devices with different computational power. Proposed another vision that consists of defining a cloud of robots and not a cloud for robots. In this perspective, we started to work with ROS which appears in our analysis as the most suitable middleware for robotic fleets.

D’Andrea [19] proposed three broader goals for the research community. The first is to collectively define the standards for — and co-develop — an indoor position system (IPS) that robots can use. Not only better visualization tools and integrated, multi-scale, and multi-physics simulation environments, although these are important. What is additionally required is a theory and calculus of information flow, decomposition, and hierarchy that takes into account reliability, robustness, and security constraints, to name but a few attributes. Design tools can be built upon such a theory and thus have a firm footing; they can, for example, provide guidelines for how a system can be decomposed, how much information should flow between components, and how much computation should reside in each subsystem. In addition, they should freely borrow from well-established system design principles such as design for verifiability, reconfigurability, and maintainability. These tools will not be fully automated: design is an art, and design without a human element is doomed to fail. Rather, they will empower robot system designers to quickly explore possible solution spaces, with the rote aspects of design delegated to algorithms. Lastly, the community should strive to provide an environment where post-PhDs can become experts at robot system design, a competency that requires both breadth and depth.

Summary

To summarize, some trends and patterns can be seen that in the topic. In theory: all RFMS include three key components: task assignment, path planning (centralized or decentralized), coordination (or traffic management). In method: Spatio-temporal trajectory envelopes for motion planning; case studies for theoretical researches and lab experiments for practical ones. In results: demonstrations in a relevant lab environment. Themes recur across the literature are the importance of RMFS connectivity and integration with other systems; fleet size significantly affects RFMS performance and may require different architecture. Debates take place on localization (positioning) type: internal or external; path planning method: centralized or decentralized; general solving paradigm: constraint satisfying or optimal. Gaps in RFMS knowledge can be highlighted: requirements, performance metrics, approaches for job (task) allocation, the line between traffic management (coordination), and motion planning, approaches for traffic management (preventing and resolving deadlocks).

References

[1] Singhal, A., Pallav, P., Kejriwal, N., Choudhury, S., Kumar, S., & Sinha, R. (2017, September). Managing a fleet of autonomous mobile robots (AMR) using cloud robotics platform. In 2017 European Conference on Mobile Robots (ECMR) (pp. 1–6). IEEE.
[2] Liaqat, A., Hutabarat, W., Tiwari, D., Tinkler, L., Harra, D., Morgan, B., … & Tiwari, A. (2019). Autonomous mobile robots in manufacturing: Highway Code development, simulation, and testing. The International Journal of Advanced Manufacturing Technology, 104(9), 4617–4628.
[3] Logistics Journal https://www.logistics-journal.de/about/intralogistics accessed 12.03.2021
[4] Huang, G. Q., Chen, M. Z., & Pan, J. (2015). Robotics in ecommerce logistics. HKIE Transactions, 22(2), 68–77.
[5] Fragapane, G., Ivanov, D., Peron, M., Sgarbossa, F., & Strandhagen, J. O. (2020). Increasing flexibility and productivity in Industry 4.0 production networks with autonomous mobile robots and smart intralogistics. Annals of operations research, 1–19.
[6] Glistau, E., & Coello Machado, N. I. (2018). Industry 4.0, logistics 4.0 and materials-Chances and solutions. In Materials Science Forum (Vol. 919, pp. 307–314). Trans Tech Publications Ltd.
[7] Malus, A., & Kozjek, D. (2020). Real-time order dispatching for a fleet of autonomous mobile robots using multi-agent reinforcement learning. CIRP annals, 69(1), 397–400.
[8] Formant blog https://formant.io/news-and-blog/2019/11/19/operations/its-time-we-put-fleet-management-to-rest-here-is-why/ accessed 12.03.2021
[9] Sanneman, L., Fourie, C., & Shah, J. A. (2020). The State of Industrial Robotics: Emerging Technologies, Challenges, and Key Research Directions. arXiv preprint arXiv:2010.14537.
[10] Wurman, P. R., D’Andrea, R., & Mountz, M. (2008). Coordinating hundreds of cooperative, autonomous vehicles in warehouses. AI magazine, 29(1), 9–9.
[11] Guizzo, E. (2008). Three engineers, hundreds of robots, one warehouse. IEEE spectrum, 45(7), 26–34.
[12] Oyekanlu, E. A., Smith, A. C., Thomas, W. P., Mulroy, G., Hitesh, D., Ramsey, M., … & Sun, D. (2020). A Review of Recent Advances in Automated Guided Vehicle Technologies: Integration Challenges and Research Areas for 5G-Based Smart Manufacturing Applications. IEEE Access, 8, 202312–202353.
[13] Andreasson, H., Bouguerra, A., Cirillo, M., Dimitrov, D. N., Driankov, D., Karlsson, L., … & Stoyanov, T. (2015). Autonomous transport vehicles: Where we are and what is missing. IEEE Robotics & Automation Magazine, 22(1), 64–75.
[14] Mansouri, M., Andreasson, H., & Pecora, F. (2015). Towards hybrid reasoning for automated industrial fleet management. In 24th International Joint Conference on Artificial Intelligence (IJCAI), Buenos Aires, Argentina, 25–31 July, 2015. AAAI Press.
[15] Yao, F., Keller, A., Ahmad, M., Ahmad, B., Harrison, R., & Colombo, A. W. (2018, July). Optimizing the scheduling of autonomous guided vehicle in a manufacturing process. In 2018 IEEE 16th International Conference on Industrial Informatics (INDIN) (pp. 264–269). IEEE.
[16] Lutz, M., Verbeek, C., & Schlegel, C. (2016, September). Towards a robot fleet for intra-logistic tasks: Combining free robot navigation with multi-robot coordination at bottlenecks. In 2016 IEEE 21st International Conference on Emerging Technologies and Factory Automation (ETFA) (pp. 1–4). IEEE.
[17] Čech, M., Wicher, P., Lenort, R., Malčic, T., David, J., Holman, D., … & Záruba, J. (2020). Autonomous mobile robot technology for supplying assembly lines in the automotive industry. Acta logistica, 7(2), 103–109.
[18] Chitic, S. G., Ponge, J., & Simonin, O. (2014, October). Are middlewares ready for multi-robots systems?. In International Conference on Simulation, Modeling, and Programming for Autonomous Robots (pp. 279–290). Springer, Cham.
[19] D’Andrea, R. (2012). Guest editorial: A revolution in the warehouse: A retrospective on kiva systems and the grand challenges ahead. IEEE Transactions on Automation Science and Engineering, 9(4), 638–639.