Deep Reinforcement Learning zur Steigerung von Energieeffizienz und Pünktlichkeit von Straßenbahnen

Abstract

This work investigates how the energy efficiency and punctuality of streetcars can be increased by using AI. The AI is trained on two scenarios at three traffic times each. The determined driving profiles are compared with those of drivers from regular passenger operation as well as with a theoretical optimum determined by Dynamic Programming. In addition, transfer learning capabilities of the AI will be investigated.