交通情境模擬對自駕車發展的助力和挑戰

Benefits and Challenges of Traffic Scenario Simulation for Self-Driving Car Development

胡振煇 ； 吳柏富

《機械工業雜誌》 445期 (2020/04) Pp. 54-61

摘要

自駕車的開發目前正如火如荼地進行，已有許多公司大量開發測試Level 4以上的自駕車功能。自駕車必須針對各種不同的交通路況自行決策並且安全行駛，如何驗證日益複雜的自駕車功能是一大難題。決策控制模組是其中確保安全的關鍵。交通情境模擬可以在自駕車正式路測前針對各種可能的交通狀況驗證自駕車的決策控制。如何建立一個平台能夠彈性且貼近真實地模擬多樣多變的交通情境是一大挑戰。針對台灣汽機車、自行車及行人高度混流的交通型態加上許多狹窄巷弄的複雜路況，我們提出情境模擬測試框架，是一種省時、經濟且安全的方法，可以進一步保障自駕車和其它用路人。

關鍵字

交通情境模擬 ； 系統功能測試驗證 ； 功能安全 ； 自動駕駛 ； 自駕車 ； 決策控制

並列摘要

The autonomous driving vehicle is being developed rapidly and widely. In recent years, a number of companies are developing and testing self-driving features above Level-4 autonomous driving capabilities. Autonomous vehicles have to drive safely with decision-making under varying traffic conditions. How to verify the increasingly complicated self-driving functions is a critical problem. In particular, the decision-making and control modules are of the most important modules to ensure safety. Traffic scenario simulation can be used to verify the decision-making and control module under various traffic situations before road tests and deployments. Establishing a simulation platform that could flexibly and closely simulate diverse and changing traffic situations is a challenge. Aiming at highly mixed traffic conditions, evolving cars, motorcycles, bicycles, and pedestrians, and complex traffic situations on the narrow roads in Taiwan, we propose a scenario simulation test that offers a time-saving, economical, and safe approach to evaluate the performance of autonomous vehicle systems.

並列關鍵字

Traffic scenario simulation ； System and function validation ； Functional safety ； Autonomous driving ； Self-driving car ； Decision making and control