Modern semiconductor production is a operating at huge complexity, often involving over 1,500 products moving through 300 processing steps across 1,200 different machines. Traditional centralized scheduling methods frequently fail to optimize these NP-hard environments in real-time. To address this, the SwarmIn project leverages Artificial Bee Colony (ABC) algorithms to create a self-organizing, bottom-up scheduling system.
Unlike standard scheduling that calculates a global plan from the top down, the SwarmIn approach models machines and products as individual agents. These agents follow local rules inspired by honeybee foraging behavior. In nature, bees use a “waggle dance” to communicate food quality; in SwarmIn, product “lots” (bees) use stigmergy to leave quality information at machines (food sources), helping incoming lots choose the most efficient processing path.
Two specialized swarm variants …
The project evaluates two distinct algorithmic variants designed for different production goals:
- Variant 1: Decentralized speed. This model prioritizes feeding batch-processing machines. By reorganizing local queues to favor lots that can complete a full batch quickly, it aims to boost overall production velocity.
- Variant 2: Predictive stability. Developed to address the queue imbalances of the first variant, this version uses a “wait table” to forecast processing times. It prioritizes lots with the minimum predicted “accumulated wait time,” effectively maintaining system balance and preventing the formation of long, disruptive queues.
… achieve real-world impact.
Testing within the SwarmFabSim framework—a simulator modeling factories like Infineon Technologies on an abstract, agent-based level—revealed that these bio-inspired strategies outperform traditional first-in-first-out methods. While Variant 1 is effective for pure speed, Variant 2 shows a remarkable 28 % improvement in flow factor (the ratio of total time to processing time) and significantly reduced product tardiness. By shifting from rigid central control to a flexible swarm of agents, Lakeside Labs’s approach provides a scalable solution for the dynamic and customized production needs of the future.
Publications
K. Youssefi, M. Gojkovic, and M. Schranz: Artificial Bee Colony Algorithm: Bottom-Up Variants for the Job-Shop Scheduling Problem. In Proc. International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH), 2024.
M. Umlauft, M. Schranz, W. Elmenreich: SwarmFabSim: A Simulation Framework for Bottom-up Optimization in Flexible Job-Shop Scheduling using NetLogo. In Proc. International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH), 2022.