Chapter

Training und Auswertung

December 2019

DOI:10.1007/978-3-658-28886-0_5

In book: Steuerung eines autonomen Fahrzeugs durch Deep Reinforcement Learning (pp.47-69)

Authors:

Andreas Folkers

Atlas Elektronik

Dieses Kapitel fasst die numerischen Ergebnisse der Steuerung eines autonomen Fahrzeugs durch einen Reinforcement Learning Agenten in Form eines Neuronalen Netzes zusammen. Dazu werden zunächst in Abschnitt 5.1 das in dieser Arbeit eingeführte Trainingsverfahren ausgewertet sowie Variationen in dessen Parametrisierung betrachtet. Anschließend wird in Abschnitt 5.2 die Qualität des resultierenden Reglers auf Basis von simulierten Daten untersucht.

ResearchGate has not been able to resolve any citations for this publication.

ResearchGate has not been able to resolve any references for this publication.

Article

Energy optimization of electric vehicle's acceleration process based on reinforcement learning

November 2019 · Journal of Cleaner Production

Under the situation of unmanned driving, the energy consumption in an electric vehicle’s acceleration process can be reduced by controlling the driving behavior. So in this paper, a pedal control strategy which could optimize the energy consumption of electric vehicle’s acceleration process is proposed. The strategy is generated by the training results of reinforcement learning framework and the ... [Show full abstract] specific method of building such framework is discussed in details. Based on the training results of Q-learning-based algorithm, the relationship between the proportion of energy consumption reduction and vehicle’s acceleration time is analyzed, which illustrates the energy-saving potential of the algorithm. In order to improve the control effect of the strategy, an updated algorithm framework based on Deep Q-learning (DQN) is proposed and an improved pedal’s control strategy is obtained. Compared with the strategy obtained by Q-learning-based algorithm, the improved strategy not only achieves the same energy-saving effect, but also guarantees the stability of control effect, which is more suitable for actual use.

Preprint

Towards an Autonomous Test Driver: High-Performance Driver Modeling via Reinforcement Learning

December 2024

Success in racing requires a unique combination of vehicle setup, understanding of the racetrack, and human expertise. Since building and testing many different vehicle configurations in the real world is prohibitively expensive, high-fidelity simulation is a critical part of racecar development. However, testing different vehicle configurations still requires expert human input in order to ... [Show full abstract] evaluate their performance on different racetracks. In this work, we present the first steps towards an autonomous test driver, trained using deep reinforcement learning, capable of evaluating changes in vehicle setup on racing performance while driving at the level of the best human drivers. In addition, the autonomous driver model can be tuned to exhibit more human-like behavioral patterns by incorporating imitation learning into the RL training process. This extension permits the possibility of driver-specific vehicle setup optimization.

Preprint

Anti-Jerk On-Ramp Merging Using Deep Reinforcement Learning

September 2019

Deep Reinforcement Learning (DRL) is used here for decentralized decision-making and longitudinal control for high-speed on-ramp merging. The DRL environment state includes the states of five vehicles: the merging vehicle, along with two preceding and two following vehicles when the merging vehicle is or is projected on the main road. The control action is the acceleration of the merging vehicle. ... [Show full abstract] Deep Deterministic Policy Gradient (DDPG) is the DRL algorithm for training to output continuous control actions. We investigated the relationship between collision avoidance for safety and jerk minimization for passenger comfort in the multi-objective reward function by obtaining the Pareto front. We found that, with a small jerk penalty in the multi-objective reward function, the vehicle jerk could be reduced by 73% compared with no jerk penalty while the collision rate was maintained at zero. Regardless of the jerk penalty, the merging vehicle exhibited human-like decision-making strategies such as merging ahead or behind a main-road vehicle.

Conference Paper

Full-text available

A Flexible Vehicle Routing Reinforcement Learning Environment for the Reusability of Trained Agents

August 2024

Conference Paper

Adaptive Eco-driving of Fuel Cell Vehicles Based on Multi-light Trained Deep Reinforcement Learning

October 2021

Last Updated: 21 Nov 2024