Researchers have developed DAROM, a novel Delay-Aware Reinforcement Learning framework designed to improve autonomous driving control in highway on-ramp merging scenarios. This framework specifically addresses the challenges posed by stochastic communication delays inherent in vehicle-to-infrastructure (V2I) systems, which can degrade the performance of standard reinforcement learning agents. DAROM utilizes a Delay-Aware Encoder to infer the current state despite delayed and partially observable information, and incorporates a physics-based safety controller to mitigate collision risks. Experiments in the SUMO simulator, using data from the NGSIM dataset, show that DAROM significantly outperforms existing RL baselines, achieving over 99% success in high-density traffic with delays up to 2.0 seconds. AI
IMPACT Enhances the robustness of autonomous driving systems to communication latency, potentially improving safety and efficiency in connected vehicle environments.
RANK_REASON The cluster contains a research paper detailing a new methodology for autonomous driving. [lever_c_demoted from research: ic=1 ai=1.0]
- autonomous driving
- gated recurrent unit (GRU)
- highway on-ramp merging
- Next Generation Simulation (NGSIM) dataset
- Reinforcement Learning
- Simulation of Urban MObility (SUMO)
- vehicle-to-infrastructure (V2I) links
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
