Linear and nonlinear Model Predictive Control (MPC) for regulating pedestrian flows with discrete speed instructions

Abstract

Airports, shopping malls, stadiums, and large venues in general, can become congested and chaotic at peak times or in emergency situations. Linear Model Predictive Control (MPC) is an effective technology in generating dynamic speed or distance instructions for regulating pedestrian flows, and constitutes a promising interventional technique to improve safety and evacuation time during emergency egress operations. We compare linear and nonlinear MPC controllers and study the influence of using continuous vs. discrete control actions. We aim to evaluate the efficacy of simple instructions that pedestrians can easily follow during evacuations. Linear and Nonlinear AutoRegressive eXogenous models (ARX and NLARX) for prediction are identified from input?output data from strategically designed microscopic evacuation simulations. A microscopic simulation framework is used to design and validate different MPC controllers tuned and refined using the identified models. We evaluate the prediction models? performance and study how the controlled variable type, density, or crowd-pressure, influences the controllers? performance. As a relevant contribution, we show that MPC control with discrete instructions is ideally suited to design and deploy practical pedestrian flow control systems. We found that an adequate size of the set of speed instructions is critical to obtain a good balance between controllability and performance, and that density output control is preferred over crowd-pressure.