A Case Study on Granularity of Industrial Vision Services

Software engineering paradigms such as service-oriented architectures are increasingly often applied in the field of factory automation. Functions like robot motion planning or object recognition are provided by cloud services. A crucial architectural aspect is the granularity, i.e. the scope and size of individual services. In our case study, we examine a service-based object recognition application for a robotic assembly use case. We implement three different granularity levels, measure their communication and computation times and discuss further architectural features. The fine-granular approach encapsulates individual image processing operations as services, which have high reusability but impose large communication overheads. The medium granularity approach is object-wise and offers best reuse efficiency and cohesion. The coarse solution offers the best performance.