The simulation scan comparison process for monitoring manufacturing environments

Automated manufacturing systems have developed significantly in the last few years, but they still remain fragile and unreliable under operational conditions. The objective of this research is to develop a decision support system that can identify problems that occur in a manufacturing environment by comparing a simulation of a manufacturing process with a three dimensional (3D) scan of the physical environment. The Simulation Scan Comparison (SSC) process was designed to compare a simulation of a manufacturing process to a 3D scan of an environment. The SSC process integrates three components: Environmental Analysis, Task Analysis and Path Planning. The Environmental Analysis is based on using a simulation for background subtraction of the 3D scan. This technique is capable of identifying known simulation objects, detecting missing simulation objects and isolating unknown objects from the scan data. The Environmental Analysis was tested by analysing the background subtraction process. For this analysis the scan was divided into areas where the scan matched to objects, areas conflicted with simulation objects and areas not matched to any simulation objects. Three scenarios were created to test the matching process: The simulation and scan environment were kept the same to test the matching of known objects; objects were added to the simulation but not the physical environment to test the missing object detection; objects were added to the physical environment, but not the simulation environment, to test the unknown object detection. The analysis of these scenarios showed that the SSC process could successfully detect that simulation objects were correctly matched 98% of the time and that unknown objects were present 90% of the time. The use of oriented bounding boxes to measure unknown objects was analysed to determine the reliability of the measurements produced. This shows that the measurement reliability of the objects was related to the scan resolution and that to be detected successfully objects should be at least three scan resolutions in any dimension. Two tracking algorithms based on analysing the unknown object data from unknown object were tested; a position based method and a velocity based method. The results demonstrate that the SSC object isolation technique can be utilised to identify and track unknown objects in a known workspace. This demonstrates that the unknown object detections are robust enough over time to perform object tracking.