Incorporating A Priori Statistical Situational Knowledge into Model Predictive Control Algorithms

Zenodo (CERN European Organization for Nuclear Research)

Abstract

In this paper, a priori situational knowledge about motion cueing for recorded test drives is incorporated into a model predictive motion cueing algorithm to move a driving simulator with greater foresight. By training a neural network on offline computed Oracle MPC motion cueing sequences, data-driven references are obtained that guide the real-time-capable MPC algorithm. The effectiveness of this modified MPC algorithm is evaluated in emergency braking scenarios near intersections by comparing the cueing error to a baseline.

Authors 5

1. Philippe Legran Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

2. Fabian Ruß Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

3. Jobst Nikolaus Bertram Beckmann Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

4. Sven Tarlowski Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

5. Lutz Eckstein