| 제목 |
관측 노이즈에 강건한 V2I 기반 xLSTM 운전 성향 추론 모델 연구
|
| 분야 |
ITS/차량 소프트웨어 |
| 언어 |
Korean |
| 저자 |
박상은(숭실대학교), 엄찬인(숭실대학교), 권민혜(숭실대학교) |
| Key Words |
Autonomous driving system(자율주행 시스템), Driving characteristics inference(운전 성향 추론), Extended long short-term memory(확장된 장단기 기억 모델), Trajectory data(궤적 데이터), Noise robustness(노이즈 강건성), Vehicle-to-infrastructure(차량-인프라 간 통신) |
| 초록 |
With the advancement of autonomous driving systems, achieving smoother interactions between autonomous and non-autonomous vehicles have become increasingly crucial. This can be achieved by inferring the driving characteristics of adjacent vehicles and incorporating them into the decision-making processes. Recently, the use of vehicle-to- infrastructure (V2I) has increased, enabling roadside units (RSUs) to collect driving information. This study proposes a model to infer driving characteristics by using trajectory data gathered through RSUs. Specifically, the inference model is based on Extended Long Short-Term Memory (xLSTM), and considers a driving environment with observational noise that may occur during the sensing process. Simulation results show that the proposed model outperformed traditional vehicle-based data collection method in complex environments, achieving accuracy. Furthermore, the RSU-driven approach demonstrates robust performance amid the presence of sensing noise and real-world driving datasets, thus highlighting its practicality. |
| 원문(PDF) |
다운로드
|
