Design and development of a context-aware collaborative autonomous real-time vehicle systems framework

Abstract

The increasing autonomy of intelligent systems, with applications extending from self-driving vehicles to home-based robots, has emerged as a critical area of focus in modern research. Yet, to acknowledge the full potential of these systems, numerous challenges must be addressed. This thesis encapsulates rigorous research resulting in eight scientific papers investigating autonomous systems’ efficacy and efficiency. Our study proposes the Context-Aware Collaborative Autonomous Real-Time Vehicle Systems (CARVS) Framework and focuses on improving context awareness, simplifying remote task processing, and quantifying prediction uncertainty in Machine Learning (ML) algorithms. Our intention is to move forward the state-of-the-art in autonomous systems based on our findings as we investigate the employment of noise as a stimulus to boost agent exploration. We also address the development of mapping and task management systems for connected autonomous vehicles (CAVs) using edge, fog, and cloud computing. Furthermore, we study the quantification of uncertainty in ML algorithm predictions to describe their behaviours and decision-making mechanisms. This research provides valuable insights for the continuous improvement of autonomous learning and the ability to deal with uncertainties in dynamic and unpredictable environments, which could lead to greater acceptance of such systems.