AI models struggle with noisy data for automated driving classification

Researchers have evaluated the effectiveness of three sequence-based models—GRU, LSTM, and Transformer encoder models—for classifying automated driving systems using vehicle telematics data. All models demonstrated strong performance on clean data, achieving macro F1-scores above 0.90. However, when subjected to realistic telematics degradation, particularly temporal jitter, the models' performance significantly dropped, with macro F1-scores falling to between 0.44 and 0.50. AI

IMPACT Highlights the critical need for robust AI models in safety-sensitive applications like automated driving, especially when dealing with imperfect data.

RANK_REASON Academic paper detailing a sensitivity analysis of AI models for a specific application. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Cadillac Super Cruise
• Comma Openpilot
• gated recurrent unit
• long short-term memory
• OpenPilot
• Tesla Autopilot
• Transformer encoder model

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →