Instance Segmentation to Path Planning in a Simulated Industrial Environment

This paper presents an integrated framework that combines instance segmentation and path planning in a simulated industrial environment to improve mobile robot navigation. By training the Mask-RCNN model using a synthetic dataset, accurate object detection and segmentation are achieved. The resulting segmentation masks are used to generate an Occupancy Grid Map (OGM) that provides detailed information about the dynamic environment. Utilizing the OGM in the path planning process enables the mobile robots to make informed decisions, considering the presence and characteristics of objects. This allows for efficient path optimization and obstacle avoidance. The framework's effectiveness is evaluated through extensive experiments, comparing it to traditional path planning approaches. The results demonstrate significant improvements in navigation performance, with the integrated framework achieving higher path efficiency and reduced travel time. Real-time instance segmentation enhances the system's adaptability to changes in the environment. The proposed framework contributes to advancing intelligent automation in industrial settings, enabling safe and efficient robot navigation. Overall, the integration of instance segmentation and path planning through the proposed framework offers a valuable solution for enhancing mobile robot navigation in industrial environments, optimizing productivity, and facilitating seamless human-robot collaboration.