rollinghorizon

Rolling horizon, also known as rolling-horizon planning, is a planning and optimization technique used in operations research and management science. In this approach decisions are optimized over a finite planning window that moves forward in time as new information becomes available. The method blends forecasting with optimization to continually adapt plans.

At time t, an optimization model is solved for periods t through t plus the planning horizon H to minimize costs, subject to constraints such as demand, capacity, inventory balance, and lead times. Only the actions for the current period are implemented, after which the horizon is shifted forward (rolled) and the process is repeated with updated forecasts and actual results. This creates an iterative loop that accommodates new data and changing conditions.

Applications span a wide range of domains, including production planning, inventory management, supply chain design, workforce scheduling, energy systems, and project scheduling. Rolling horizon approaches are particularly valuable when demand and supply conditions are dynamic and forecasts can be improved over time.

The method is closely related to receding horizon control and model predictive control, where optimization over a moving horizon informs immediate decisions. In stochastic or robust variants, uncertainty is explicitly incorporated through probabilistic models or worst-case considerations.

Advantages include better responsiveness to new information, more flexible use of up-to-date data, and the ability to avoid overly long-term commitments. Limitations involve computational burden, dependence on forecast quality, potential myopia if the horizon is too short, and possible instability if re-optimization induces frequent changes in decisions.

Effective use requires choosing an appropriate horizon length and update frequency, along with reliable data integration to feed forecasts and actual results into the optimization model.