Optimising robotic pick-and-place operations and the application of RMS principles

Robotic Pick-and-Place (PnP) kitting operations involve transferring components from storage locations to assembly positions under strict component-requirement compatibility constraints. In this work, this setting is formulated as a mapping-constrained Bipartite Travelling Salesman Problem (BTSP), systematically examined under three different industrially-driven graph cases. An Integer Programming (IP)-based exact solution method together with adjusted heuristics and reinforcement learning approaches are employed to generate feasible and optimal PnP sequences by explicitly enforcing compatibility and operational constraints. Notably, the IP approach achieves execution time reductions of up to 23.80% over the industrial baseline. To further improve system performance, a simulation-driven framework is introduced, where storage configuration is treated as a decision parameter and evaluated through repeated reconstruction of BTSP instances. This enables systematic exploration of layout alternatives, producing additional efficiency gains of 13.6% on average and up to 35.73%. The proposed framework is deployed within a Digital Twin-enabled architecture and validated in a real industrial environment, demonstrating how optimisation and simulation support the practical application of Reconfigurable Manufacturing Systems principles.