The key to reduce unreasonable risk due to a hazard, namely SOTIF requirement, is to evaluate the autonomous driving performance when facing known hazardous scenarios. Autonomous driving performance will be largely optimized by testing their evasion and mitigation ability in reproduced hazardous scenario. China Automotive Engineering Research Institute established an automatic transferring tool which transfer the raw scenario data to OpenX scenario. And we developed a parallel simulation platform based on naturalistic driving data, which successfully evaluate the autonomous driving performance automatically and support the iterative development. This report will detailed illustrate techniques about automatic scenario transfer, parallel simulation and track testing and autonomous driving performance evaluation.