RoboPlanner

Abstract

Autonomous robots are being increasingly integrated into Industry 4.0 manufacturing and retail industries due to the twin advantages of improved throughput and adaptivity. In order to handle complex tasks, the autonomous robots require robust action plans, that can self-adapt to runtime changes. A further requirement is efficient implementation of knowledge bases, that may be queried during planning and execution. In this paper, we propose RoboPlanner, a framework to generate action plans in autonomous robots. In RoboPlanner, we model the knowledge of world models, robotic capabilities and task templates using knowledge property graphs and graph databases. Design time queries and robotic perception are used to enable intelligent action planning, that can adapt at runtime. We demonstrate these solutions on autonomous picker robots deployed in Industry 4.0 warehouses.