Self-organization is a powerful method to bring forward automation in industrial applications. However, tasks in such use cases often have high complexity and strong interdependencies, limiting the applicability of state-of-the-art self-organization algorithms. This paper studies self-organization algorithms that are suitable for task allocation in industrial manufacturing processes with strong task dependencies. Two algorithms based on the Response Threshold Model are proposed in this paper to fit task dependencies. For decision-making, they rely solely on local interactions between the involved agents and the environment. To analyze their performance, an upper bound of the achievable performance is calculated using methods from Synchronous Data Flow. Simulations are performed under both homogeneous and heterogeneous agents and different task-switching costs. The results indicate that the DRTM FR-based algorithm, which evolves agents into a specialized team, has a better performance.