<p>With the rapid advancement of intelligent manufacturing, multi-robot collaborative systems are becoming increasingly integrated into complex production environments. This study addresses a multi-robot collaborative flexible job shop scheduling problem (FJSP) characterized by controllable processing time coupled with time-window constraints. This combination yields a complex NP-hard problem that poses significant challenges for existing methods. To tackle these challenges, this study develops two mixed integer linear programming (MILP) models, a constraint programming (CP) model, and a hybrid genetic algorithm-constraint programming (GA-CP) method. Two novel strategies are incorporated to enhance the GA-CP method: a dynamic iteration control strategy and an adaptive elimination mechanism for filtering low-potential solutions. Extensive computational experiments are conducted on 96 benchmark cases and 10 real-world cases to validate the proposed approach. Results demonstrate that MILP models provide accurate solutions for small-scale problems but become computationally intractable as problem size increases. In contrast, the GA-CP method identifies best-known solutions in 75 benchmark cases and consistently outperforms standalone MILP and CP models. Compared with recent adaptive large neighborhood search algorithmic framework with tabu-based component and without an optimal repair strategy (ALNST) and variable neighborhood search (VNS) methods, the GA-CP shows superior performance, particularly on medium-scale and large-scale cases. Additionally, small-scale dynamic experiments involving new order insertions and machine breakdowns demonstrate that the proposed GA-CP can serve as a foundation for future dynamic scheduling extensions. These findings validate that combining global search with constraint-based local optimization represents an effective and scalable strategy for complex scheduling problems in aircraft skin components manufacturing with multi-robot collaboration and time-window constraints.</p>