A multi-level intelligent optimisation framework for the efficient implementation of human-robot collaborative assembly systems

<p dir="ltr">In the context of Industry 5.0, assembly systems are evolving to emphasise humancentricity, flexibility, and resilience. Central to this transformation is the rise of collaborative robots (cobots), which reshape human-machine interactions by enabling flexible task sharing and synergistic cooperation on human-robot collaborative assembly lines (HRAL), integrating human cognitive adaptability with robotic precision and endurance. However, despite the promising advantages of HRAL, a critical question remains unanswered: how can manufacturing industries effectively integrate human and cobot capabilities to overcome practical implementation challenges and achieve sustained performance gains across different decision-making horizons? Firstly, at the strategic level, selecting and configuring suitable cobot systems is complex and lacks comprehensive decision-making tools tailored to real-world operational conditions. Secondly, at the tactical level, efficiently scheduling tasks and balancing workloads among heterogeneous human-robot teams is problematic, as current scheduling methods insufficiently consider workforce diversity. Lastly, at the operational level, industries struggle with productivity losses caused by inherent variations in human skill levels and dynamic uncertainties during real-time human-robot interactions, significantly impacting operational efficiency and incurring substantial economic costs. To systematically address these interconnected challenges, this thesis proposes a multilevel intelligent optimisation framework specifically designed for the efficient implementation of HRAL. At the strategic level, the framework introduces an integrated multi-criteria decision-making (MCDM) approach that combines a novel "Triple-I" taxonomy: spanning technical, integration, and organisational barriers, with the Best- Worst Method (BWM) and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) for systematic evaluation and ranking of cobot configurations. At the tactical level, a novel optimisation model is developed for the Assembly Line Worker Integration and Balancing Problem with multi-skilled Human-Robot Collaboration (ALWIBP-mHRC), explicitly accounting for workforce heterogeneity and diverse collaboration modes. This tactical challenge is further addressed by the design of an Adaptive Multi-Objective Cooperative Co-evolutionary algorithm (a-MOCC). The a- MOCC algorithm decomposes the scheduling problem and employs adaptive evolutionary operators, enabling the discovery of high-quality task allocations that maximise productivity while minimising the cobot investment cost. At the operational level, a hybrid optimisation–simulation environment, the Metaheuristic-driven Agent-Based Simulation Environment (MASE), is proposed. MASE integrates tactical-level optimisation results with an agent-based digital twin of the HRAL to evaluate real-time system behaviour under uncertainty. By simulating production with intelligent human and cobot agents under dynamic conditions (e.g. fluctuating worker performance or potential human-robot conflicts), this approach identifies robust production plans that minimise productivity losses and sustain high throughput despite variability. Collectively, the proposed multi-level framework creates an integrated decision-support pipeline guiding manufacturers from initial cobot adoption decisions through detailed production planning to real-time operational control. The significance of this framework extends beyond operational performance, directly contributing to the Industry 5.0 paradigm. Ergonomically, it assigns physically strenuous tasks to robots, enhancing worker well-being. Socially, it supports inclusivity by enabling the integration of semiskilled or disadvantaged workers. Economically and environmentally, it develops resilience and robustness against variability and disruptions, sustaining productivity and reducing resource waste. Ultimately, this thesis advances both the technical capabilities and sustainability goals of modern manufacturing, paving the way for a future-ready human-robot collaborative assembly paradigm.</p>