A Korean AI Model Scores Driving Paths for Safety
CVPR called it a highlight.
SafeDrive: A Novel Approach to Autonomous Driving Safety
SafeDrive, developed by a team at Seoul National University led by professor Jun Won Choi, generates multiple driving trajectories, scores each for safety using sensor data, and selects the best path. This innovative technique, termed Fine-grained Safety Reasoning, sets a new standard in autonomous driving safety.
The model stands out as the first Korean end-to-end autonomous driving paper to earn a CVPR highlight, signifying South Korea's growing presence in this field, traditionally dominated by US and Chinese research labs.
Addressing Self-Driving AI Limitations
Most self-driving AIs mimic human driving styles, performing well under normal conditions but struggling to explain their path choices, which can prove problematic in split-second decisions. SafeDrive offers a solution by providing transparency and auditable decision trails, crucial for regulatory, insurance, and legal purposes when autonomous vehicles make mistakes.
Integration and Commercialization
The technique has already been integrated into EAD, a reference model backed by Korea's Ministry of Trade, Industry, and Energy. SafeDrive is currently being tested in real vehicles with domestic autonomous driving companies, paving the way for commercialization using proprietary driving data.
The Future of Autonomous Driving
With significant investment in AI, chips, and robotics, South Korea's SafeDrive model represents a tangible step forward in making autonomous driving safer and more explainable. As Tesla robotaxis face challenges with crash rates, models like SafeDrive offer promising solutions to enhance safety and public trust in this emerging technology.