Safe Model Predictive Control via Reliable Time-Series Forecasting

The control of dynamical systems is the backbone of modern technologies, ranging from industrial processes to autonomous vehicles. In many of these scenarios, systems must be controlled while satisfying a set of safety and reliability constraints concerning the unknown evolution of a target process.

In our recent work[1], we proposed Probabilistic Time Series-Conformal Risk Prediction (PTS-CRC), a novel calibration procedure that enables reliable modelling of uncertainty regarding future system states. PTS-CRC predictions can be used to solve model predictive control problems under reliability, safety, and performance constraints.

Consider the UAV tracking problem illustrated in Figure 1. Here, a UAV must plan its trajectory to track a group of ground users with unknown future locations. In this example, PTS-CRC predictions can replace the unknown future locations of the users and assist in designing a UAV trajectory that minimizes energy consumption while ensuring reliability in tracking errors.

Figure 1: UAV tracking problem, an example of model predictive control in which the UAV must plan its path based on the unknown evolution of the object to be tracked. PST-CRC-based model predictive control replaces the unknown future location of the tracking target with reliable time series prediction sets.

[1] Zecchin, Matteo, Sangwoo Park, and Osvaldo Simeone. “Forking Uncertainties: Reliable Prediction and Model Predictive Control with Sequence Models via Conformal Risk Control.” arXiv preprint arXiv:2310.10299, 2023.