Model-driven allocation engineering: specifying and solving constraints based on the example of automotive systems

Automotive systems provide sophisticated functionality and are controlled by networked electronic control units (ECUs). Nowadays, software engineers use component-based development approaches to develop their software. Moreover, software components have to be allocated to ECUs to be executed. Engineers have to cope with topology-, software-, and timing dependencies and memory-, scheduling-, and routing constraints. Currently, engineers use linear programming to specify allocation constraints manually and to compute a feasible allocation specification automatically. However, encoding the allocation problem as a linear program is a complex and error-prone task. This paper contributes a model-driven, object constraint language based, and graph pattern based allocation engineering approach for reducing the engineering effort and to avoid failures. We validate our approach with an automotive case study. We specify the software component model, the hardware platform model, and the allocation constraint specification with our engineering approach MechatronicUML. Our validation shows that we can specify allocation constraints with less engineering effort and are able to compute feasible allocation specifications automatically.