The objective of this thesis is to establish a Closed-Loop Supply Chain (CLSC) design that is analysed through a series of simulation models, aimed at defining the highest performing production control strategy, whilst considering multiple related variables on both the forward and reverse flow of materials in manufacturing environments. Due to its stochastic nature, the reverse logistics side of the CLSC represents an increased source of variance for the inventory management and control strategies as it implies the erratic supply of returned materials, in addition to the very random customer demand, hence with highly variable inputs on both sides of the productive system, intrinsically inherent to this line of research. To test the operational performance of several pull-type production control strategies, a simulation-based research method was designed. The strategies experimented were: Hybrid Extended Kanban CONWIP special case (HEKC-II), Hybrid Kanban CONWIP (HKC), Dynamic Allocation Hybrid Extended Kanban CONWIP special case (DNC HEKC-II) and Dynamic Allocation Hybrid Kanban CONWIP (DNC HKC). All were tested in scenarios with high and low processing time variability and with 90% returned products and 40% returns from an open market system, therefore totaling 16 simulation models. Multi-objective evolutionary algorithms were utilised to generate the Pareto-optimum performance frontier with the objective of simultaneously minimising both performance metrics: The overall average work in progress (WIP) and the average backlog queue length (BL) for the entire CLSC. Processes used in the recovery and recycling of end of life manufactured goods were examined. This research method structures leading factors towards improved economic viability and sustainability of technologies required for the effective implementation of inventory control strategies on highly complex closed-loop supply chains with the focus on the performance metrics and optimum utilisation of resources available for the industry. The dynamic allocation strategies proved significant performance improvement, shifting the entire Pareto frontier forward with major advances on both metrics. Furthermore, it happened on all scenarios tested. The modified HEKC-II, with an optimisable parameter that enables it to be overwritten in a way that it can match the well-established HKC, also performed as originally intended and had better results than HKC in some cases, especially with the higher variability level. It also provided grounds for the suggested improvements and flexibilisation of the HEKC strategy. A major contribution of this thesis was the successful implementation of another advanced control methodology, entitled here the Intelligent Self-Designing Production Control Strategy, which provided maximum control performance. It consisted essentially of DNC HEKC-II with the following modifications: I) Extensive increase of dynamically allocated authorisation cards; II) Further anticipation of the time to trigger the change in the number of cards according to the finished goods buffer level, plus an acceleration/deceleration factor of this change; III) The capability of downsizing itself to become similar to HKC in an optimisation process if diverse production system conditions and variability would require. It displayed a very significant shift of the performance frontier.