Building an End-User Development Framework for the Internet of Things

Abstract

The proliferation of Internet of Things (IoT) devices is evident in various settings, ranging from monitoring production processes in factories to providing customised lighting solutions in homes. However, as the number of IoT devices continues to increase, a critical question is how to program these devices to meet the expectations of an expanding user base. This user base increasingly expects devices to exhibit smart behaviour and enable ambient intelligence, which poses a significant challenge to professional developers who can not cope with this growing demand. Against this backdrop, this thesis explores methodologies for constructing IoT workflows that require minimal intervention from professional developers. Here, the term workflow alludes to the sequence of tasks each device should perform in a given timeline. The thesis demonstrates that the expectations placed on IoT workflows continue to rise, highlighting the need for innovative solutions to meet this demand. In particular, the thesis discusses the requirements of IoT to support culinary processes. The culinary domain is favoured for this thesis since it is ubiquitous---it provides multiple instructional texts (i.e. online cooking recipes), ample knowledgeable users exist in the cooking domain, and users are expected to have high engagement with the IoT. Consequently, it presents an ideal domain for testing new methodologies for constructing IoT workflows with minimal professional developer involvement. Two main methodologies are investigated towards constructing the IoT workflow for the least involvement of professional developers: 1. The explicit involvement of end users in developing the workflow is coined as end-user development (EUD). Specifically, the thesis analyses the meta-design EUD principle to allow multiple end users with varying degrees of code competency to collaborate towards constructing workflows. The vision envisages various EUD programming styles integrated for altering IoT workflows. 2. The thesis explored how specifically, natural language can act as a EUD to develop IoT workflows; under the premise that the user can submit a recipe as a program to orchestrate their IoT kitchen devices. For that matter, experiments were accomplished on information extraction from instructional texts (i.e. cooking recipes) to facilitate the IoT workflow construction. Moreover, it investigated how domain experts can directly optimise the natural language processing (NLP) models through interactive machine learning. The results indicate that EUD meta-design could facilitate various end users, and NLP can extract the required information for the IoT workflow.

Date of Award	1 Aug 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Sarah Clinch (Supervisor) & Riza Theresa Batista-Navarro (Supervisor)