Grant: Discovery Grants Program - Individual

Internet of Things (IoT) systems are prevalent nowadays, with an estimated 15 billion connected devices in 2023. They combine networks of connected devices with software systems that produce information and control physical systems. The components of IoT systems include small devices, e.g., microcontrollers running C programs, Edge devices, e.g., Raspberry Pis running Python scripts, and Cloud servers, e.g., Amazon EMR running Apache Spark. They communicate via dedicated or standard protocols, e.g., MQTT or HTTPs. Developers use these components and protocols via IoT framework APIs (APIs for short). However, (1) these APIs are used in a variety of contexts and by a diversity of users, often not software developers. (2) They have grown organically from cyber-physical systems, with little systematic standardisation and software-quality consideration. (3) They are driven by the capabilities of the components to which they give access, e.g., types of available sensors. While standards exist, e.g., for communication protocols, many APIs give access to similar capabilities, each in their own way, while others are unique to some components. Only a few APIs apply to several, different components. This variety and heterogeneity in APIs confuse IoT developers, hinder choosing and using IoT components and APIs, and increase vendor and technology lock-ins. To help IoT developers choose and use IoT components, we must bridge developers' requirements, those fulfilled by IoT frameworks, and the IoT framework APIs and guide developers to explore the large space of all these requirements and APIs. My pioneering research program will relate the requirements of IoT components with the specification and usage of IoT framework APIs through the features and patterns realising these requirements. It is original by studying the relations among and the characteristics of the requirements, features, and patterns of IoT APIs (1) to understand the (dis)similarities among requirements and APIs. It is innovative by producing recommendation systems built on these relations and characteristics (2) to replace the paradigm of random exploration with one of evidence-based recommendation. For each of my objectives, I will follow multi-method approaches in a combination of qualitative and quantitative analyses. Qualitatively, I will perform systematic literature reviews to collect APIs requirements, features, patterns, and their quality characteristics. I will survey and interview students and professionals to validate the results of the reviews as well as, later, the appropriateness and usefulness of the recommendations. Quantitatively, I will use natural language processing techniques, static and dynamic analyses, as well as historical analysis to collect relevant data from Q&A Web sites, to trace requirements, features, and patterns, to analyse existing APIs and identify different patterns of specification/usage, and to implement recommendation systems using these patterns.