Rollinghorisonts

Rollinghorisonts is a concept in forecasting and decision making that describes the continuous reassessment and extension of the planning horizon as new information becomes available. It treats the planning interval not as a fixed window but as a rolling target that adapts to data quality, computational resources, and changing conditions, with the goal of aligning decisions with the most relevant future.

The concept draws on rolling horizon planning and model predictive control. At each decision epoch, a finite-horizon optimization is solved over an adaptive horizon length H(t). After implementing the first control action, new observations are incorporated and the horizon is updated, so the next optimization considers a refreshed forecast over a possibly different horizon. This rolling process is designed to maintain responsiveness while preserving computational tractability.

Applications include supply chain management, energy system operation, transportation and logistics, and autonomous robots. In non-stationary environments, rollinghorisonts can help balance short-term feasibility with long-term goals by extending the horizon when forecasts are reliable and shortening it when uncertainty is high.

Challenges include selecting the update rule for H(t), ensuring stability and convergence, managing computational burden, and avoiding overfitting to recent data. Studies compare rollinghorisonts with static-horizon approaches and other adaptive planning methods to determine when and how horizon adjustments yield benefits.

Related concepts include rolling horizon planning, model predictive control, adaptive forecasting, and dynamic optimization.