2023/09/12

Can Hong Kong tap the opportunities in autonomous driving? - EJ Insight

網上版請按此

 

Can Hong Kong tap the opportunities in autonomous driving?

An essential driver for autonomous driving (AD) development is precision of maps.

There are two very different approaches for AD, the so-called Waymo model and the Tesla version, according to Wang Xiao, CTO and co-founder of Beijing AD system developer Idriverplus.

Wang explained that the Waymo model, which aimed at the highest levels (L4/5) of AD at the beginning, deploying complex and expensive sensors and equipment. The cost and technical challenges of this approach are extremely high, making it difficult to commercialise. The Tesla model (self-declared as L2) is based on continuous data collection through relatively inexpensive sensors, such as cameras mounted on mass-produced vehicles. With massive data, its automation algorithms can gradually improve.

The major difference between the two modes, in addition to the cost of sensor equipment, is whether to use high-definition maps (HD maps) or not. Serving as the "eyes" of AV, HD maps provide road geometry, route profile, and traffic signs which are essential for highly automated driving.

According to an insider of the Baidu Map, L2 vehicles mainly requires maps for advanced driver assistance systems (ADAS), with an accuracy of between 1 and 5 meters, and the map information mainly provides the basic features and geometry of the road. The precise level of HD maps is however down to centimetre, with L3 between 0.2-0.5 metres, and L4/5 as precise as 0.1 meters.

The cost of achieving such precision is high. A vehicle with high-precision map collection capability in Mainland is reported to cost more than one million yuan. The cost of a fleet of 200 vehicles, including labor and maintenance, is roughly estimated to be one billion yuan. However, this scale of investment may only be able to cover the national expressway with the mileage of about 177,000 kilometers, accounting for just 3.3% of the total mileage of highways of the country. Considering those in the urban cities with much richer details, data collection becomes a huge challenge, to maintain HD maps updated becomes even more difficult.

Today, in China, players including Internet giants, traditional map makers, overseas suppliers, domestic innovation companies, and more, have engaged in HD map making, leading to diversified data sources with different labeling methods. Due to a lack of industry standard, data cannot be shared between enterprises.

In view of this, China's Ministry of Natural Resources released a 2023 edition of guideline for the standard on intelligent-vehicle basic maps in March this year. To meet the needs of the country's technological and industrial development, it includes formulating more than 10 standards by 2025, covering technical requirements and specifications such as general application, data collection and distribution, production update, quality detection, and security management to address the urgent need for in-depth application of basic maps. By 2030, a relatively complete basic map standard for intelligent-vehicles should be finalised.

Meanwhile, investment bank Goldman Sachs expects the global HD maps market to grow to US$9.4 billion by 2025. The mainland industry also generally believes that the HD map industry will enter a golden era in the next 15 years.

With the rapid development of HD maps, the competition for talents will intensify. How can we take advantage of such great opportunities? The UK Geospatial Strategy 2030 published by the Geospatial Commission of the U.K. in mid-2023 proposes a series of talent training programmes, including the launch of a pilot programme this year to upgrade the skill of geographers in public sector with digital and coding skills, in preparation for a wider rollout to other public and private sectors. At the same time, the Commission has also supported various local universities to establish master's degree and doctoral training centers in spatial science, and plans to embed geospatial into the university's geography and data science courses in the near future... Will the Hong Kong government draw inspiration from this?

 

 

 

Dr. Winnie Tang
Adjunct Professor, Department of Computer Science, Faculty of Engineering; Department of Geography, Faculty of Social Sciences; and Faculty of Architecture, The University of Hong Kong