A type-2 fuzzy logic based goal-driven simulation for optimising field service delivery

This thesis develops an intelligent system capable of incorporating the conditions that drive operational activity while implementing the means to handle unexpected factors to protect business sustainability. This solution aims to optimise field service operations in the utility-based industry, especially within one of the world's leading communications services companies, namely BT (British Telecom), which operates in highly regulated and competitive markets. Notably, the telecommunication sector is an essential driver of economic activity. Consequently, intelligent solutions must incorporate the ability to explain their underlying algorithms that power their final decisions to humans. In this regard, this thesis studies the following research gaps: the lack of integrated solutions that go beyond isolated monolithic architectures, the lack of agile end-to-end frameworks for handling uncertainty while business targets are defined, current solutions that address target-oriented problems do not incorporate explainable methodologies; as a result, limited explainability features result in inapplicability for highly regulated industries, and most tools do not support scalability for real-world scenarios. Hence, the need for an integrated, intelligent solution to address these target-oriented simulation problems. This thesis aims to reduce the gaps mentioned above by exploiting fuzzy logic capabilities such as mimicking human thinking and handling uncertainty. Moreover, this thesis also finds support in the Explainable AI field, particularly in the strategies and characteristics to deploy more transparent intelligent solutions that humans can understand. Hence, these foundations support the thesis to unlock explainability, transparency and interpretability. This thesis develops a series of techniques with the following features: the formalisation of an end-to-end framework that dynamically learns form data, the implementation of a novel fuzzy membership correlation analysis approach to enhance performance, the development of a novel fuzzy logic-based method to evaluate the relevancy of inputs, the modelling of a robust optimisation method for operational sustainability in the telecommunications sector, the design of an agile modelling approach for scalability and consistency, the formalisation of a novel fuzzy-logic system for goal-driven simulation for achieving specific business targets before being implemented in real-life conditions, and a novel simulation environment that incorporates visual tools to enhance interpretability while moving from conventional simulation to a target-oriented model. The proposed tool was developed based on data from BT, reflecting their real-world operational conditions. The data was protected and anonymised in compliance with BT’s sharing of information regulations. The techniques presented in the development of this thesis yield significant improvements aligned to institutional targets. Precisely, as detailed in Section 9.5, the proposed system can model a reduction between 3.78% and 5.36% of footprint carbon emission due to travel times for jobs completion on customer premises for specific geographical areas. The proposed framework allows generating simulation scenarios 13 times faster than conventional approaches. As described in Section 9.6, these improvements contribute to increased productivity and customer satisfaction metrics regarding keeping appointment times, completing orders in the promised timeframe or fixing faults when agreed by an estimated 2.6%. The proposed tool allows to evaluate decisions before acting; as detailed in Section 9.7, this contributes to the ‘promoters’ minus ‘detractors’ across business units measure by an estimated 1%.