This presentation explores the evolving optimization landscape of modern Multihead Weigher Machines (MWM). As industrial demands for speed and precision increase, MWM designs have become significantly more complex, transitioning toward advanced double-layer configurations with higher hopper counts. Our initial research successfully implemented an Ant Colony Optimization (ACO) approach, achieving optimal values in 97.3% of test cases for up to 32 hoppers. However, newly engineered designs, such as those incorporating booster hoppers, introduce exponential growth in combinatorial possibilities. Our current work addresses these challenges by focusing on two critical factors: architectural complexity and real-world product variability. Although double-layer designs theoretically enhance accuracy by increasing the number of available combinations, they impose a substantial computational burden on MWM microcontrollers. In addition, the unpredictable mass distribution of non-uniform goods complicates the optimization process, often compromising the precision levels suggested by simulations. A key contribution of this ongoing research is the development of an optimization framework validated against new instances derived from physical experiments, providing a realistic representation of operational uncertainties. Additionally, we will demonstrate the deployment and benchmarking of these algorithms on resource-constrained devices, specifically the ESP32, to evaluate computational efficiency across continuous multiple weighing batches. This study offers vital insights into achieving high-precision target weights within complex, high-speed automated packaging environments under strict hardware limitations.