Recently, the second ICVS China Autonomous Driving Annual Conference in 2022 was held in Shanghai. Mingshang Technology won the 2022 Autonomous Driving Innovation Technology Award with its strong technical strength in the field of intelligent driving in commercial vehicles.
At the annual meeting, Dr. You Jie, Chief Technology Officer of Mingshang Technology, delivered a keynote speech titled “Deep Practice and Exploration of Intelligent Driving in Commercial Vehicles”, focusing on the two major technological product elements of “data engine+AI container”, sharing Mingshang Technology’s deep practice and core technological advantages in the field of intelligent driving in commercial vehicles.

Dr. You opened the keynote speech with “Can Autonomous Vehicles Think?” to stimulate the audience’s profound thinking on the essence behind autonomous driving technology and applications. Subsequently, a two-layer structural model for intelligent driving was proposed by referencing human thinking models. The “subconscious” layer, which provides the basic automation capabilities for intelligent driving (including sensing, communication, data fusion, data processing, perception, map and positioning, decision-making, control, and redundant design, etc.), is shared by passenger and commercial vehicles for intelligent driving. On the subconscious level, passenger cars and commercial vehicles need to think about different issues on the conscious level, which is the main difference between intelligent driving systems for passenger cars and commercial vehicles. Passenger cars mainly provide superior services for personal travel and entertainment, while the issues that need to be considered in the commercial vehicle field are much more complex, with different pain points and needs in different commercial scenarios. Ports are different from mines Logistics parks are different from municipal sanitation. In summary, the issues related to different business scenarios are all about quality, efficiency, and cost. However, the Know how of different business scenarios determines the uniqueness and solutions of their respective quality, efficiency, and cost.

Based on a two-layer model, Dr. You proposed Mingshang Technology’s unique approach to intelligent driving solutions for commercial vehicles: on top of the intelligent driving base that provides universal basic capabilities for autonomous driving, Mingshang Technology focuses more on solving commercial problems at the “consciousness” level, which are mainly defined by ODD (Operational Design Domain). To achieve precise and valuable intelligent driving solutions for commercial vehicles, Firstly, it is necessary to define the Operable Domain (ODD) that the solution is oriented towards. ODD includes a series of constraints, mainly including road conditions, geographical conditions, environmental conditions, and other commercial conditions. Business conditions refer to the knowledge of specific business scenarios, which are more specific and complex, but regardless of the type of conditions, they are ultimately defined by data. Whether it is the image and video data that sense the road, the LiDAR point cloud data that obtains spatial location, the ultrasonic data that detects distance, or the data dictionary and API interface that defines commercial logic. So, data determines ODD and ultimately determines intelligent driving solutions for commercial vehicles.

Dr. You said, “The combination of data closed-loop and algorithm iteration speed is the key to winning in intelligent driving of commercial vehicles.” In 20 years of commercial practice, Mingshang Technology has transformed this concept into its own product system. Based on the essence of AI intelligent technology, this project aims to meet the business needs and challenges of special business scenarios. With intelligent onboard devices as the starting point and commercial vehicle networking as the main body, it constructs a “dual drive” intelligent solution that combines a data engine and a multitasking AI model engine for commercial vehicle business scenarios.

1) Data engine

Mingshang has established strategic partnerships with leading domestic and foreign commercial vehicle companies such as Yutong, BYD, Sany, and Dongfeng. Through front-end installation, Mingshang’s digital base, intelligent in car cameras, and other series of products have been widely applied in commercial vehicle models, and through connecting to Mingshang’s vehicle networking platform, a wide range of multi scene real operation big data collection has been achieved. At the same time, by constructing a dedicated testing site for commercial vehicles in Xiamen, continuous collection of sample data is achieved to achieve long tail coverage of multi scene, multi object, multi state, and multi demand data.

2) Low cost automatic annotation technology

In order to meet the low-cost operational needs of various commercial customers, well-known businesses have abandoned expensive LiDAR equipment. Using full image technology, based on online vehicle networking and offline vehicle algorithm models, utilizing dual closed-loop data iteration to achieve semantic prediction, verification, and self annotation of image data, further cooperating with technologies such as binocular camera self coding, driver driving recording, and ultrasonic ranging to achieve a more comprehensive database construction for intelligent driving of commercial vehicles.

3) Multi task artificial intelligence model

The Mingshang Intelligent Driving Technology Laboratory is based in Qianhai, Shenzhen and Hong Kong, and has built an international AI technology team. Faced with diverse, fast updating, and special business scenarios, it has built a scalable and customizable multitasking AI algorithm engine. The laboratory has conducted in-depth research on the following aspects to build its own technical characteristics and barriers: a Multi task neural network architecture; b. Multi task model training and data selection and fusion during runtime; c. Design and optimization of loss functions.

4) Low cost multi perception fusion pre controller

In the face of diversified long tail commercial application scenarios, in addition to considering the special needs in different scenarios, it is also necessary to take into account the cost preferences of customers from different backgrounds. This requires the collection of diversified data (including ultrasound, millimeter wave, thermal imaging, video, LiDAR, etc.) to meet the demand problem, and to create an optional hardware combination item that can go from cost optimal to function optimal. In response to this, Mingshang has collaborated with vehicle intelligent chip manufacturers such as Horizon to polish and optimize a series of commercial vehicle intelligent driving pre controller products that can meet the requirements of basic monitoring to remote scheduling, basic ADAS, AEB to L4 level commercial vehicle unmanned driving, as well as special scenario requirements (such as mineral processing and loading and unloading).

As a leading provider of intelligent driving systems for commercial vehicles in the industry, Mingshang Technology will take “making driving safer and travel more convenient” as its mission, actively embrace the changing trends in the industry, adhere to technological innovation as the lead, and continue to promote the digital, networked, and information-based development of commercial vehicles.