Temporal-Aware Reinforcement Learning for Robust Autonomous Driving Decision-Making
- Autonomous driving systems require robust and precise decision-making, particularly for safety-critical functions such as Automatic Emergency Braking (AEB) and for comfortoriented features like Adaptive Cruise Control (ACC). Although conventional Reinforcement Learning (RL) methods with Multilayer Perceptrons (MLPs) have shown potential, they lack the temporal awareness needed for complex, dynamic driving scenarios. In this paper, we propose an enhanced decisionmaking architecture that integrates Proximal Policy Optimization (PPO) with sequential models—specifically Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks. Simulation-based evaluations demonstrate that our method improves AEB responsiveness and ACC stability while reducing false activations.
| Author: | Xin Xing, Jannes BikkerORCiD, Sebastian OhlORCiD |
|---|---|
| URN: | urn:nbn:de:bsz:960-opus4-37928 |
| DOI: | https://doi.org/10.25968/opus-3792 |
| Parent Title (German): | KI-Forum 2025 : KI in Forschung und Lehre an Hochschulen |
| Publisher: | HsH Applied Academics |
| Place of publication: | Hannover |
| Editor: | Hanno Homann, Cedric Rohbani, Jens Christian Will |
| Document Type: | Conference Proceeding |
| Language: | English |
| Year of Completion: | 2025 |
| Publishing Institution: | Hochschule Hannover |
| Release Date: | 2025/12/10 |
| Tag: | Autonomous Driving; Reinforcement Learning; Sequential Modeling |
| GND Keyword: | Künstliche IntelligenzGND; Autonomes FahrzeugGND; Bestärkendes Lernen <Künstliche Intelligenz>GND |
| Page Number: | 4 |
| First Page: | 117 |
| Last Page: | 120 |
| Institutes: | Sonstige Einrichtungen |
| DDC classes: | 370 Erziehung, Schul- und Bildungswesen |
| 004 Informatik | |
| 621.3 Elektrotechnik, Elektronik | |
| Licence (German): |  Creative Commons - CC BY - Namensnennung 4.0 International |
