Addressing Stochasticity in Business Process Management

Traditional business process scheduling often assumes a deterministic control flow, where the sequence of tasks is known in advance. This research addresses the reality of 'control-flow uncertainty,' where branching paths and conditional logic make the exact sequence of operations probabilistic. The authors propose a mathematical model that integrates stochastic programming with process mining data to optimize resource allocation under these conditions.

Mathematical Framework for Scheduling

The core contribution is a formulation that treats process execution as a stochastic optimization problem. By modeling the probability of different control-flow paths, the framework allows organizations to:

• Minimize expected completion time (makespan) while accounting for path variability.
• Optimize resource utilization by predicting the likelihood of task activation across different branches.
• Apply mixed-integer linear programming (MILP) techniques to solve for optimal scheduling policies that remain robust even when specific process branches are triggered.

Practical Implications for Process Automation

This approach moves beyond static scheduling by providing a dynamic way to handle uncertainty. Instead of planning for a single 'happy path,' the model generates schedules that account for the statistical distribution of process variants. This is particularly valuable for complex enterprise workflows where resource bottlenecks often occur at conditional junctions. By quantifying the impact of control-flow uncertainty, builders can design more resilient automated pipelines that adapt to real-world process variance rather than failing when execution deviates from the ideal path.