A Computational Framework for Automatic Online Path Generation of Robotic Inspection Tasks via Coverage Planning and Reinforcement Learning

Abstract

Surface/shape inspection is a common and highly repetitive task in the factory production line. Using robots to automate the inspection process could help to reduce the costs and improve the productivities. In robotized surface/shape inspection application, the planning problem is to find a near-optimal sequence of robotic actions that inspect the surface areas of the target objects in a minimum cycle time, while satisfying the coverage requirement. In this paper, we propose a novel computational framework to automatically generate efficient robotic path online for surface/shape inspection application. Within the computational framework, a Markov decision process (MDP) formulation is proposed for the coverage planning problem in the industrial surface inspection with a robotic manipulator. A reinforcement learning-based search algorithm is also proposed in the computational framework to generate planning policy online with the MDP formulation of the robotic inspection problem for robotic inspection applications. Several case studies are conducted to validate the effectiveness of the proposed method. It is observed that the proposed method could automatically generate the inspection path online for different target objects to meet the coverage requirement, with the presence of pose variation of the target object. In addition, the inspection cycle time reduction is observed to be 24% on average compared to the previous approaches during these test instances.

Keywords

• Robotics
• planning
• artificial intelligence
• reinforcement learning