A Reinforcement Learning Approach to View Planning for Automated Inspection Tasks.

Manual inspection of workpieces in highly flexible production facilities with small lot sizes is costly and less reliable compared to automated inspection systems. Reinforcement Learning (RL) offers promising, intelligent solutions for robotic inspection and manufacturing tasks. This paper presents an RL-based approach to determine a high-quality set of sensor view poses for arbitrary workpieces based on their 3D computer-aided design (CAD). The framework extends available open-source libraries and provides an interface to the Robot Operating System (ROS) for deploying any supported robot and sensor. The integration into commonly used OpenAI Gym and Baselines leads to an expandable and comparable benchmark for RL algorithms. We give a comprehensive overview of related work in the field of view planning and RL. A comparison of different RL algorithms provides a proof of concept for the framework's functionality in experimental scenarios. The obtained results exhibit a coverage ratio of up to 0.8 illustrating its potential impact and expandability. The project will be made publicly available along with this article.

Tumor size and detection in breast cancer: Self-examination and clinical breast examination are at their limit.

BACKGROUND: This study investigates to what extent tumor detection methods in breast cancer have changed and how this has influenced tumor size at initial diagnosis. METHODS: 1054 breast carcinomas < or =5 cm, newly diagnosed between 1990 and 2006, were evaluated for the tumor detection methods used, namely self-detection (SD, n=568), clinical breast examination (CBE, n=212), and radiological breast examination (RBE, n=237), and their corresponding tumor sizes. RESULTS: During the study period, the proportion of cases found by RBE increased (p<0.001), while median tumor size decreased (1990-1992: 22 mm; 2005/2006: 17 mm. Spearman rho=-0.12, p<0.001). Nevertheless, SD remained the most frequent method of tumor identification (2005/2006: 48.9%). Carcinomas found by RBE were smaller (median size: 12 mm) than those found by the other two detection forms (SD: 21 mm, CBE: 21 mm; p<0.001). Within the different methods, only in RBE was an appreciable decrease in the size of the detected tumors observed during the study period (Spearman rho=-0.14, p<0.001; SD: Spearman rho=-0.05, p=0.19; CBE: Spearman rho=-0.05, p=0.43). CONCLUSION: Despite educational campaigns and high media coverage, the possibilities for improving the "classical" methods of tumor detection in breast cancer, self-detection and clinical breast examination, seem to be at their limit. The significant decrease in tumor size at time of detection observed in the last years is primarily only due to the increased use of breast imaging. Improved detection of smaller tumors may presumably be reached only by an increased use of radiological procedures.