Engineering Heterogeneous Internet of Things Applications: From Models to Code

Abstract

Applications on top of the Internet of Things (IoT) show big potential, but the domain raises challenges for software engineers. Heterogeneous hardware environments and software stacks, unreliable devices, and diverse developer skillsets make the implementation of business processes spanning the entire application and the enforcement of constraints difficult. To this end, we propose a methodology and technical framework targeting heterogeneous IoT systems, where software components may be situated on IoT devices, cloud infrastructure, or edge devices - a paradigm often referred to as fog computing. Our approach leverages design-time modeling of device interfaces, data, and business processes. Design-time activities entail specification of the software architecture of the IoT application in an implementation-and language-agnostic manner. Subsequently, we automatically generate boilerplate code for participating devices, requiring developers to only implement business processes. The generated boilerplate code is correct by construction and it targets the different languages usually involved in diverse IoT software components. When the application is deployed, its execution may violate design assumptions. To counter this, constraints defined during design time are enforced at runtime, ensuring that devices operate within certain bounds. We evaluate our model-driven engineering framework over a health care system use case.