Network-based order grouping and resource planning joint optimisation under digitalised customisation

Digitalised customisation requires firms to fulfil stochastically arriving and structurally heterogeneous product–service orders under limited resources. In this setting, conventional similarity-based grouping becomes increasingly inadequate. This paper addresses the problem from two perspectives. First, it develops a network-based order grouping approach that jointly considers scale economies and scope economies to improve order aggregation under random and diverse order arrivals. Second, recognising the coupling between order allocation and resource integration, it proposes a novel two-timescale stochastic optimisation model that integrates decisions on resource integration and order allocation. The slow timescale determines resource-integration decisions and the fast-timescale scheduling problem is formulated as a Markov decision process for the dynamic allocation and scheduling of aggregated orders. On this basis, an Actor-Critic method is developed to learn adaptive scheduling policies under stochastic arrivals and evolving structural states. Numerical results show that the proposed framework consistently outperforms conventional benchmarks, with stronger advantages under higher demand intensity and order heterogeneity.