Towards Autonomous Intralogistics: A Testbed Environment for the Coordination of a Robotic Fleet
In Proceedings of the International Conference on Automated Planning and Scheduling (ICAPS)---Workshop on Planning and Robotics (PlanRob), (ICAPS-2020), 20.10.-20.10.2020, Online, ICAPS, Oct/2020.
In this paper we present a testbed implementation, in whicha production plant for airplane wings is simulated. The simulated environment can be used for the evaluation of different a pproaches to multi-agent task allocation and path finding. We outline the relation to the relevant problems of job-shop scheduling and multi-agent path finding and describe our approach
towards a system architecture and a baseline implementation of the required software components for a multiagent system to tackle these challenges. Our approach facilitates a centralized database where the state of the production environment is consolidated (a so called ”digital twin”), a high-level planner that continuouslydispatches the available agents to tasks, and a multi-agent
simulation to simulate the task execution. In general our system is designed in a modular fashion, which allows the developer to extend or replace the distinct components easily. The modular design also makes it possible to factor out the multi-robot simulation in favor of a real-world test environment with actual robotic agents. We describe the relevant parameters that are tracked about theenvironment, the design of the planner, and the interactions between the planner and the agents, that are executing the plans autonomously. We present first experiments conducted within this environment as a starting point for a series of runs of experimentation.