한국자동차연구원(이하 한자연)이 자율주행 업계에서는 HD Map의 필요성에 대해 상반된 입장이 공존하고 있으며 이는 산업의 발전을 위해 제작 활용 측면에서 새로운 접근 방식이 요구됨을 시사했다.
HD Map, real-time autonomous driving decision-making, sensor misrecognition prevention features
Tesla, “HD Map, Difficulty in Reflecting Road Changes in Real Time and High Cost Burden”
The Korea Automobile Research Institute (KARI) reported that there are conflicting views on the need for HD Maps in the autonomous driving industry, suggesting that a new approach is required in terms of production and utilization for the advancement of the industry.
In the Industry Trend Vol. 119 published under the theme of 'Autonomous Driving Related HD Map Issues and Implications', Hanjayeon analyzed that the controversy is likely to continue as the claims of the autonomous driving HD map, led by Tesla and Huawei, and the HD map production industry are in conflict.
HD Map contains detailed information in 3D, including lane units, and even utilizes the functions of sensors that expand the range of perception in autonomous driving, so specialized companies such as HERE and TomTom and big tech companies such as Mobileye and Waymo are jumping into the related business or forming partnerships.
The HD Map is composed of information such as road type, width, stop zone, and speed, and is used for real-time autonomous driving decision-making, and is composed of information such as traffic signal signs, and is used to quickly and accurately identify the vehicle location on the map, and is evaluated to have great advantages in autonomous driving.
However, the disadvantage of producing an HD map is that it requires a lot of resources as special vehicles are driven on real roads.
According to Hanjayeon's report, the price of a vehicle equipped with a high-performance system is estimated to be around 1 billion won per unit, and the vehicle is also When operated, a lot of costs are incurred in maintaining the vehicles, and a lot of resources are consumed in maintaining and expanding coverage as MMS vehicles must repeatedly drive over already mapped areas for updates.
Tesla creates a vector space consisting of a narrow map of the surrounding environment and surrounding vehicles based on real-time video data collected from vehicle cameras, and implements autonomous driving based on this.
Since the road environment is constantly changing, it is difficult to produce and use high-precision maps in advance. We are approaching this from the perspective that vehicles should be able to instantly identify and judge the environment like human drivers by using data collected by sensors.
Based on this approach, autonomous driving can always be implemented based on the latest information, so the pre-construction of an HD map is unnecessary.
Opinions are growing over the need for precise maps as China's Huawei also announced plans to launch a solution that does not rely on HD Maps by 2023, following Telsa's policy.
On the other hand, the HD Map industry recognizes HD Map as an essential element for fully autonomous driving.
The claim is that if the constructed HD Map is utilized, the current location can be identified with high accuracy even in bad weather or GPS shadow areas where the sensor signal is weak or likely to be misrecognized, and if the HD Map is dedicated to all environmental resources, the computing burden can also be reduced.
The industry is trying out crowdsourcing-based mapping, edge computing, and AI-based automation to make up for the aforementioned shortcomings, and is conducting research to streamline costs and time.
Han Ja-yeon said, “The position that it is difficult to exclude the use of HD Maps in fully autonomous driving is still intact, so the need for HD Maps is a key factor in autonomous driving. He said, “The controversy is likely to continue as it is closely related to the viewpoint on the conditions for realization,” and “In order for the related industry to develop, it is necessary to resolve the shortcomings and expand the application of the business.”