Abstract
Modern shop floors are evolving into activity-based environments, emphasising flexibility and efficiency in task allocation. This thesis investigates predictive modeling, reward systems, and decision-making processes to enhance task management in such dynamic shop floor settings. Machine learning models are employed to predict task occurrences using operational data, addressing the challenge of fluctuating number of tasks through weekly task occurrence distributions that incorporate worker availability and historical trends. A simulation framework compared static and dynamic reward systems, demonstrating the ability of dynamic systems to adapt to real-time task acceptance rates, reducing unaccepted tasks. Building on this, a reward model incorporating social cooperation is introduced, leveraging a benefit-cost mechanism to represent team reputation. The approach significantly reduced task delays, improved machine utilisation, and highlighted its potential to minimise energy consumption and emissions when applied with full cooperation strategies. Recognising the influence of cognitive heuristics on decision-making, an experimental study in ecological shopping provided analogies to worker behaviours on shop floors. It revealed the impact of local and cost heuristics on decisions, even when recommendations were provided, emphasising the need to address these heuristics in recommendation system design. This work advances activity-based management systems, offering practical insights for implementing adaptive reward models, fostering cooperation, and mitigating the negative impact of decision-making heuristics in modern manufacturing environments.